Evaluation of variable selection methods for random forests and omics data sets

Models of ultrasonic radiomics and clinical characters for lymph node metastasis assessment in thyroid cancer: a retrospective study

Background

Preoperative prediction of cervical lymph node metastasis in papillary thyroid carcinoma provided a basis for tumor staging and treatment decision. This study aimed to investigate the utility of machine learning and develop different models to preoperatively predict cervical lymph node metastasis based on ultrasonic radiomic features and clinical characteristics in papillary thyroid carcinoma nodules.

Methods

Data from 400 papillary thyroid carcinoma nodules were included and divided into training and validation group. With the help of machine learning, clinical characteristics and ultrasonic radiomic features were extracted and selected using randomforest and least absolute shrinkage and selection operator regression before classified by five classifiers. Finally, 10 models were built and their area under the receiver operating characteristic curve, accuracy, sensitivity, specificity, positive predictive value and negative predictive value were measured.

Results

Among the 10 models, RF-RF model revealed the highest area under curve (0.812) and accuracy (0.7542) in validation group. The top 10 variables of it included age, seven textural features, one shape feature and one first-order feature, in which eight were high-dimensional features.

Conclusions

RF-RF model showed the best predictive performance for cervical lymph node metastasis. And the importance features selected by it highlighted the unique role of higher-dimensional statistical methods for radiomics analysis.

Changes in arsenic accumulation and metabolic capacity after environmental management measures in mining area

Due to the public health concern of arsenic, environmental management measures in mining areas had been implemented. To assess the effect of environmental management measures in the mining area comprehensively, arsenic accumulation in the urine, hair, nails, and urinary metabolites of residents in a realgar mining area in Hunan province, China were investigated in 2019, and the changes in arsenic levels in the biomarkers during 2012-2019 were tracked. The importance of confounding factors (age, sex, occupation, residence, clinical history, vegetable source, cooking fuel, smoking, alcohol consumption, BMI) was analyzed using the Boruta algorithm. After the implementation of environmental management measures (including ceasing mining and smelting activities, building landfills, adjusting the planting structure, and soil restoration), urine, hair, and nail arsenic concentration decreased drastically but were still excessive. Arsenic accumulation was highest in older male miners who were long settled in the mining area and consumed homegrown vegetables. The only factor for changes in urinary arsenic levels was the cooking fuel type; residents using wood as cooking fuel experienced sustained arsenic exposure. Occupation and sex were important for determining arsenic changes in the hair and nails. Short-term arsenic accumulation in urine was affected by arsenic exposure, while long-term accumulation in hair and nails by arsenic metabolic capacity. The percentage of urinary arsenic metabolism and arsenic methylation indices of the participants in the mining area were within the normal range (%iAs: 10-30 %, %MMA: 10-20 %, % DMA: 60-80 %); samples indicated worse metabolic capacity than the reference population. The arsenic metabolic capacity of male miners was relatively weak, probably aggravated by alcohol drinking and smoking. Without soil remediation, arsenic exposure will continue. Homegrown vegetables and biomass fuels should be abandoned; reduced cigarette and alcohol consumption is recommended. Urinary arsenic would be more proper for assessing environmental remediation in mining areas.

A Novel Online Nomogram Established with Five Features before Surgical Resection for Predicating Prognosis of Neuroblastoma Children: A Population-Based Study

Background: Neuroblastoma (NB) is the most common childhood cancer, but doctors are unable to predict its overall survival (OS) rate before surgery. We aimed to predict the OS of NB children with some clinical features obtained from biopsy before surgery. Methods: Clinical features of NB children were retrospectively collected from the Therapeutically Applicable Research to Generate Effective Treatments database. The C-index, area under the receiver operating characteristic curve (AUC), calibration curves, and decision curves analysis were used to estimate nomogram models. Results: A total of 488 NB children were evaluated, and the Boruta algorithm was used to detect risk factors. The results showed that artificial neural networks with selected features were able to predict more than 90% of NB children. Five risk factors were used in the construction of the nomogram, including age at diagnosis, MYCN status, ploidy value, histology, and mitosis-karyorrhexis index (MKI). The C-index of the nomogram in training cohort and validation cohort was 0.716 and 0.731. AUC values for 1-, 3-, and 5-years OS predictions were 0.706, 0.755, and 0.762, respectively, and showed good calibrations. Decision curve analysis indicated a better predictability with the nomogram model based on Cox regression compared with one that included all variables and histology only. Also, the Kaplan-Meier curves showed a significantly higher survival probability in the low-risk group (total score <118.34) versus the high-risk group (total score ≥ 118.34) (p < 0.05) using the nomogram model. Conclusions: A web application based on the nomogram model in the present study can be accessed at https://mdzhou.shinyapps.io/DynNomapp/, which could help doctors make accurate clinical decisions about NB children.

Design and methodological considerations for biomarker discovery and validation in the Integrative Analysis of Lung Cancer Etiology and Risk (INTEGRAL) Program

The Integrative Analysis of Lung Cancer Etiology and Risk (INTEGRAL) program is an NCI-funded initiative with an objective to develop tools to optimize low-dose CT (LDCT) lung cancer screening. Here, we describe the rationale and design for the Risk Biomarker and Nodule Malignancy projects within INTEGRAL. The overarching goal of these projects is to systematically investigate circulating protein markers to include on a panel for use (i) pre-LDCT, to identify people likely to benefit from screening, and (ii) post-LDCT, to differentiate benign versus malignant nodules. To identify informative proteins, the Risk Biomarker project measured 1161 proteins in a nested-case control study within 2 prospective cohorts (n = 252 lung cancer cases and 252 controls) and replicated associations for a subset of proteins in 4 cohorts (n = 479 cases and 479 controls). Eligible participants had a current or former history of smoking and cases were diagnosed up to 3 years following blood draw. The Nodule Malignancy project measured 1078 proteins among participants with a heavy smoking history within four LDCT screening studies (n = 425 cases diagnosed up to 5 years following blood draw, 430 benign-nodule controls, and 398 nodule-free controls). The INTEGRAL panel will enable absolute quantification of 21 proteins. We will evaluate its performance in the Risk Biomarker project using a case-cohort study including 14 cohorts (n = 1696 cases and 2926 subcohort representatives), and in the Nodule Malignancy project within five LDCT screening studies (n = 675 cases, 680 benign-nodule controls, and 648 nodule-free controls). Future progress to advance lung cancer early detection biomarkers will require carefully designed validation, translational, and comparative studies.

Predicting pathological highly invasive lung cancer from preoperative [18F]FDG PET/CT with multiple machine learning models

PURPOSE

The efficacy of sublobar resection of primary lung cancer have been proven in recent years. However, sublobar resection for highly invasive lung cancer increases local recurrence. We developed and validated multiple machine learning models predicting pathological invasiveness of lung cancer based on preoperative [¹⁸F]fluorodeoxyglucose (FDG) positron emission tomography (PET) and computed tomography (CT) radiomic features.

METHODS

Overall, 873 patients who underwent lobectomy or segmentectomy for primary lung cancer were enrolled. Radiomics features were extracted from preoperative PET/CT images with the PyRadiomics package. Seven machine learning models and an ensemble of all models (ENS) were evaluated after 100 iterations. In addition, the probability of highly invasive lung cancer was calculated in a nested cross-validation to assess the calibration plot and clinical usefulness and to compare to consolidation tumour ratio (CTR) on CT images, one of the generally used diagnostic criteria.

RESULTS

In the training set, when PET and CT features were combined, all models achieved an area under the curve (AUC) of ≥ 0.880. In the test set, ENS showed the highest mean AUC of 0.880 and smallest standard deviation of 0.0165, and when the cutoff was 0.5, accuracy of 0.804, F1 of 0.851, precision of 0.821, and recall of 0.885. In the nested cross-validation, the AUC of 0.882 (95% CI: 0.860-0.905) showed a high discriminative ability, and the calibration plot indicated consistency with a Brier score of 0.131. A decision curve analysis showed that the ENS was valid with a threshold probability ranging from 3 to 98%. Accuracy showed an improvement of more than 8% over the CTR.

CONCLUSION

The machine learning model based on preoperative [¹⁸F]FDG PET/CT images was able to predict pathological highly invasive lung cancer with high discriminative ability and stability. The calibration plot showed good consistency, suggesting its usefulness in quantitative risk assessment.

Identification of biomarkers to stratify response to B-cell-targeted therapies in systemic lupus erythematosus: an exploratory analysis of a randomised controlled trial

Background

Systemic lupus erythematosus (SLE) is a complex autoimmune disease associated with widespread immune dysregulation and diverse clinical features. Immune abnormalities might be differentially associated with specific organ involvement or response to targeted therapies. We aimed to identify biomarkers of response to belimumab after rituximab to facilitate a personalised approach to therapy.

Methods

In this exploratory analysis of a randomised controlled trial (BEAT-LUPUS), we investigated immune profiles of patients with SLE recruited to the 52-week clinical trial, which tested the combination of rituximab plus belimumab versus rituximab plus placebo. We used machine learning and conventional statistics to investigate relevant laboratory and clinical biomarkers associated with major clinical response. BEAT LUPUS is registered at ISRCTN, 47873003, and is now complete.

Findings

Between Feb 2, 2017, and March 28, 2019, 52 patients were recruited to BEAT-LUPUS, of whom 44 provided clinical data at week 52 and were included in this analysis. 21 (48%) of 44 participants were in the belimumab group (mean age 39·5 years [SD 12·1]; 17 [81%] were female, four [19%] were male, 13 [62%] were White) and 23 (52%) were in the placebo group (mean age 42·1 years [SD 10·5]; 21 [91%] were female, two [9%] were male, 16 [70%] were White). Ten (48%) of 21 participants who received belimumab after rituximab and eight (35%) of 23 who received placebo after rituximab had a major clinical response at 52 weeks (between-group difference of 13% [95% CI -15 to 38]). We found a predictive association between baseline serum IgA2 anti-double stranded DNA (dsDNA) antibody concentrations and clinical response to belimumab after rituximab, with a between-group difference in major clinical response of 48% (95% CI 10 to 70) in patients with elevated baseline serum IgA2 anti-dsDNA antibody concentrations. Moreover, among those who had a major clinical response, serum IgA2 anti-dsDNA antibody concentrations significantly decreased from baseline only in the belimumab group. Increased circulating IgA2 (but not total) plasmablast numbers, and T follicular helper cell numbers predicted clinical response and were both reduced only in patients who responded to belimumab after rituximab. Serum IgA2 anti-dsDNA antibody concentrations were also associated with active renal disease, whereas serum IgA1 anti-dsDNA antibody and IFN-α concentrations were associated with mucocutaneous disease activity but did not predict response to B-cell targeted therapy. Patients with a high baseline serum interleukin-6 concentration were less likely to have a major clinical response, irrespective of therapy.

Interpretation

This exploratory study revealed the presence of distinct molecular networks associated with renal and mucocutaneous involvement, and response to B-cell-targeted therapies, which, if confirmed, could guide precision targeting of advanced therapies for this heterogenous disease.

Funding

Versus Arthritis, UCLH Biomedical Research Centre, LUPUS UK, and GSK.

Risk Stratification of COVID-19 Using Routine Laboratory Tests: A Machine Learning Approach

The COVID-19 pandemic placed significant stress on an already overburdened health system. The diagnosis was based on detection of a positive RT-PCR test, which may be delayed when there is peak demand for testing. Rapid risk stratification of high-risk patients allows for the prioritization of resources for patient care. The study aims were to classify patients as severe or not severe based on outcomes using machine learning on routine laboratory tests. Data were extracted for all individuals who had at least one SARS-CoV-2 PCR test conducted via the NHLS between the periods of 1 March 2020 to 7 July 2020. Exclusion criteria: those 18 years, and those with indeterminate PCR tests. Results for 15437 patients (3301 positive and 12,136 negative) were used to fit six machine learning models, namely the logistic regression (LR) (the base model), decision trees (DT), random forest (RF), extreme gradient boosting (XGB), convolutional neural network (CNN) and self-normalising neural network (SNN). Model development was carried out by splitting the data into training and testing set of a ratio 70:30, together with a 10-fold cross-validation re-sampling technique. For risk stratification, admission to high care or ICU was the outcome for severe disease. Performance of the models varied: sensitivity was best for RF at 75% and accuracy of 75% for CNN. The area under the curve ranged from 57% for CNN to 75% for RF. RF and SNN were the best-performing models. Machine Learning (ML) can be incorporated into the laboratory information system and offers promise for early identification and risk stratification of COVID-19 patients, particularly in areas of resource-poor settings.

Predicting Length of Stay using machine learning for total joint replacements performed at a rural community hospital

BACKGROUND

Predicting patient's Length of Stay (LOS) before total joint replacement (TJR) surgery is vital for hospitals to optimally manage costs and resources. Many hospitals including in rural areas use publicly available models such as National Surgical Quality Improvement Program (NSQIP) calculator which, unfortunately, performs suboptimally when predicting LOS for TJR procedures.

OBJECTIVE

The objective of this research was to develop a Machine Learning (ML) model to predict LOS for TJR procedures performed at a Perioperative Surgical Home implemented rural community hospital for better accuracy and interpretation than the NSQIP calculator.

METHODS

A total of 158 TJR patients were collected and analyzed from a rural community hospital located in Montana. A random forest (RF) model was used to predict patient's LOS. For interpretation, permuted feature importance and partial dependence plot methods were used to identify the important variables and their relationship with the LOS.

RESULTS

The root mean square error for the RF model (0.7) was lower than the NSQIP calculator (1.21). The five most important variables for predicting LOS were BMI, Duke Activity Status-Index, diabetes, patient's household income, and patient's age.

CONCLUSION

This pilot study is the first of its kind to develop an ML model to predict LOS for TJR procedures that were performed at a small-scale rural community hospital. This pilot study contributes an approach for rural hospitals, making them more independent by developing their own predictions instead of relying on public models.

Metabolomics and random forests in patients with complex congenital heart disease

Introduction

It is increasingly common to simultaneously determine a large number of metabolites in order to assess the metabolic state of, or clarify biochemical pathways in, an organism (“metabolomics”). This approach is increasingly used in the investigation of the development of heart failure. Recently, the first reports with respect to a metabolomic approach for the assessment of patients with complex congenital heart disease have been published. Classical statistical analysis of such data is challenging.

Objective

This study aims to present an alternative to classical statistics with respect to identifying relevant metabolites in a classification task and numerically estimating their relative impact.

Methods

Data from two metabolomic studies on 20 patients with complex congenital heart disease and Fontan circulation and 20 controls were reanalysed using random forest (RF) methodology. Results were compared to those of classical statistics.

Results

RF analysis required no elaborate data pre-processing. The ranking of the variables with respect to classification impact (subject diseased, or not) was remarkably similar irrespective of the evaluation method used, leading to identical clinical interpretation.

Conclusion

In metabolomic classification in adult patients with complex congenital heart disease, RF analysis as a one-step method delivers the most adequate results with minimum effort. RF may serve as an adjunct to traditional statistics also in this small but crucial-to-monitor patient group.

Newborn Cry Acoustics in the Assessment of Neonatal Opioid Withdrawal Syndrome Using Machine Learning

IMPORTANCE

The assessment of opioid withdrawal in the neonate, or neonatal opioid withdrawal syndrome (NOWS), is problematic because current assessment methods are based on subjective observer ratings. Crying is a distinctive component of NOWS assessment tools and can be measured objectively using acoustic analysis.

OBJECTIVE

To evaluate the feasibility of using newborn cry acoustics (acoustics referring to the physical properties of sound) as an objective biobehavioral marker of NOWS.

DESIGN, SETTING, AND PARTICIPANTS

This prospective controlled cohort study assessed whether acoustic analysis of neonate cries could predict which infants would receive pharmacological treatment for NOWS. A total of 177 full-term neonates exposed and not exposed to opioids were recruited from Women & Infants Hospital of Rhode Island between August 8, 2016, and March 18, 2020. Cry recordings were processed for 118 neonates, and 65 neonates were included in the final analyses. Neonates exposed to opioids were monitored for signs of NOWS using the Finnegan Neonatal Abstinence Scoring Tool administered every 3 hours as part of a 5-day observation period during which audio was recorded continuously to capture crying. Crying of healthy neonates was recorded before hospital discharge during routine handling (eg, diaper changes).

EXPOSURES

The primary exposure was prenatal opioid exposure as determined by maternal receipt of medication-assisted treatment with methadone or buprenorphine.

MAIN OUTCOMES AND MEASURES

Neonates were stratified by prenatal opioid exposure and receipt of pharmacological treatment for NOWS before discharge from the hospital. In total, 775 hours of audio were collected and trimmed into 2.5 hours of usable cries, then acoustically analyzed (using 2 separate acoustic analyzers). Cross-validated supervised machine learning methods (combining the Boruta algorithm and a random forest classifier) were used to identify relevant acoustic parameters and predict pharmacological treatment for NOWS.

RESULTS

Final analyses included 65 neonates (mean [SD] gestational age at birth, 36.6 [1.1] weeks; 36 [55.4%] female; 50 [76.9%] White) with usable cry recordings. Of those, 19 neonates received pharmacological treatment for NOWS, 7 neonates were exposed to opioids but did not receive pharmacological treatment for NOWS, and 39 healthy neonates were not exposed to opioids. The mean of the predictions of random forest classifiers predicted receipt of pharmacological treatment for NOWS with high diagnostic accuracy (area under the curve, 0.90 [95% CI, 0.83-0.98]; accuracy, 0.85 [95% CI, 0.74-0.92]; sensitivity, 0.89 [95% CI, 0.67-0.99]; specificity, 0.83 [95% CI, 0.69-0.92]).

CONCLUSIONS AND RELEVANCE

In this study, newborn acoustic cry analysis had potential as an objective measure of opioid withdrawal. These findings suggest that acoustic cry analysis using machine learning could improve the assessment, diagnosis, and management of NOWS and facilitate standardized care for these infants.

Interpretable Machine Learning-Based Prediction of Intraoperative Cerebrospinal Fluid Leakage in Endoscopic Transsphenoidal Pituitary Surgery: A Pilot Study

Purpose  Transsphenoidal surgery (TSS) for pituitary adenomas can be complicated by the occurrence of intraoperative cerebrospinal fluid (CSF) leakage (IOL). IOL significantly affects the course of surgery predisposing to the development of postoperative CSF leakage, a major source of morbidity and mortality in the postoperative period. The authors trained and internally validated the Random Forest (RF) prediction model to preoperatively identify patients at high risk for IOL. A locally interpretable model-agnostic explanations (LIME) algorithm is employed to elucidate the main drivers behind each machine learning (ML) model prediction. Methods  The data of 210 patients who underwent TSS were collected; first, risk factors for IOL were identified via conventional statistical methods (multivariable logistic regression). Then, the authors trained, optimized, and audited a RF prediction model. Results  IOL reported in 45 patients (21.5%). The recursive feature selection algorithm identified the following variables as the most significant determinants of IOL: Knosp's grade, sellar Hardy's grade, suprasellar Hardy's grade, tumor diameter (on X, Y, and Z axes), intercarotid distance, and secreting status (nonfunctioning and growth hormone [GH] secreting). Leveraging the predictive values of these variables, the RF prediction model achieved an area under the curve (AUC) of 0.83 (95% confidence interval [CI]: 0.78; 0.86), significantly outperforming the multivariable logistic regression model (AUC = 0.63). Conclusion  A RF model that reliably identifies patients at risk for IOL was successfully trained and internally validated. ML-based prediction models can predict events that were previously judged nearly unpredictable; their deployment in clinical practice may result in improved patient care and reduced postoperative morbidity and healthcare costs.

Impact of Phenol-Enriched Olive Oils on Serum Metabonome and Its Relationship with Cardiometabolic Parameters: A Randomized, Double-Blind, Cross-Over, Controlled Trial

Phenol-rich foods consumption such as virgin olive oil (VOO) has been shown to have beneficial effects on cardiovascular diseases. The broader biochemical impact of VOO and phenol-enriched OOs remains, however, unclear. A randomized, double-blind, cross-over, controlled trial was performed with thirty-three hypercholesterolemic individuals who ingested for 3-weeks (25 mL/day): (1) an OO enriched with its own olive oil phenolic compounds (PCs) (500 ppm; FOO); (2) an OO enriched with its own olive oil PCs (250 ppm) plus thyme PCs (250 ppm; FOOT); and (3) a VOO with low phenolic content (80 ppm). Serum lipid and glycemic profiles, serum ¹H-NMR spectroscopy-based metabolomics, endothelial function, blood pressure, and cardiovascular risk were measured. We combined OPLS-DA with machine learning modelling to identify metabolites discrimination of the treatment groups. Both phenol-enriched OO interventions decreased the levels of glutamine, creatinine, creatine, dimethylamine, and histidine in comparison to VOO one. In addition, FOOT decreased the plasma levels of glycine and DMSO2 compared to VOO, while FOO decreased the circulating alanine concentrations but increased the plasma levels of acetone and 3-HB compared to VOO. Based on these findings, phenol-enriched OOs were shown to result in a favorable shift in the circulating metabolic phenotype, inducing a reduction in metabolites associated with cardiometabolic diseases.

Magnetic Resonance Elastography as Surrogate Marker of Interstitial Fibrosis in Kidney Transplantation: A Prospective Study

Background

Fibrosis progression is a major prognosis factor in kidney transplantation. Its assessment requires an allograft biopsy, which remains an invasive procedure at risk of complications.

Methods

We assessed renal stiffness by magnetic resonance elastography (MRE) as a surrogate marker of fibrosis in a prospective cohort of kidney transplant recipients compared with the histologic gold standard. Interstitial fibrosis was evaluated by three methods: the semi-quantitative Banff ci score, a visual quantitative evaluation by a pathologist, and a computer-assisted quantitative evaluation. MRE-derived stiffness was assessed at the superior, median, and inferior poles of the allograft.

Results

We initially enrolled 73 patients, but only 55 had measurements of their allograft stiffness by MRE before an allograft biopsy. There was no significant correlation between MRE-derived stiffness at the biopsy site and the ci score (ρ=-0.25, P=0.06) or with the two quantitative assessments (pathologist: ρ=-0.25, P=0.07; computer assisted: ρ=-0.21, P=0.12). We observed negative correlations between the stiffness of both the biopsy site and the whole allograft, with either the glomerulosclerosis percentage (ρ=-0.32, P=0.02 and ρ=-0.31, P=0.02, respectively) and the overall nephron fibrosis percentage, defined as the mean of the percentages of glomerulosclerosis and interstitial fibrosis (ρ=-0.30, P=0.02 and ρ=-0.28, P=0.04, respectively). At patient level, mean MRE-derived stiffness was similar across the three poles of the allograft (±0.25 kPa). However, a high variability of mean stiffness was found between patients, suggesting a strong influence of confounding factors. Finally, no significant correlation was found between mean MRE-derived stiffness and the slope of eGFR (P=0.08).

Conclusions

MRE-derived stiffness does not directly reflect the extent of fibrosis in kidney transplantation.

Older Adults' concerns regarding Hurricane-Induced evacuations during COVID-19: Questionnaire findings

The COVID-19 pandemic has drastically affected our day-to-day life in the last few years. This problem becomes even more challenging when older adults are considered due to their less powerful immune system and vulnerability to infectious diseases, especially in Florida where 4.5 million people aged 65 and over reside. With its long coastline, large and rapidly growing of older adult population, and geographic diversity, Florida is also uniquely vulnerable to hurricanes, which significantly increases the associated risks of COVID-19 even further. This study investigates older adults' evacuation-related concerns during COVID-19 using statistical analysis of a questionnaire conducted among 389 older adult Florida residents. The questionnaire includes questions concerning demographic information and older adults' attitudes toward hurricane-induced evacuations during the COVID-19 pandemic. Ordered Probit regression models were developed to investigate the impacts of demographic parameters on older adults' tendencies toward evacuating as well as their preferences to stay at home or shelter during the pandemic. The model results reveal that male participants felt safer to evacuate compared to females. Also, any decrease in the level of income was associated with an increase in the need for help for evacuation by 18%. Findings indicated that the participants who found the evacuation safe normally also had a positive attitude toward staying in their vehicle, hotel, or even shelters if maintaining social distance was possible. Emergency management policies can utilize these findings to enhance hurricane preparations for dealing with the additional health risks posed by the pandemic for older adults, a situation that could be exacerbated by the upcoming hurricane season in Florida.

Ultrasound-based radiomics XGBoost model to assess the risk of central cervical lymph node metastasis in patients with papillary thyroid carcinoma: Individual application of SHAP

Objectives

A radiomics-based explainable eXtreme Gradient Boosting (XGBoost) model was developed to predict central cervical lymph node metastasis (CCLNM) in patients with papillary thyroid carcinoma (PTC), including positive and negative effects.

Methods

A total of 587 PTC patients admitted at Binzhou Medical University Hospital from 2017 to 2021 were analyzed retrospectively. The patients were randomized into the training and test cohorts with an 8:2 ratio. Radiomics features were extracted from ultrasound images of the primary PTC lesions. The minimum redundancy maximum relevance algorithm and the least absolute shrinkage and selection operator regression were used to select CCLNM positively-related features and radiomics scores were constructed. Clinical features, ultrasound features, and radiomics score were screened out by the Boruta algorithm, and the XGBoost model was constructed from these characteristics. SHapley Additive exPlanations (SHAP) was used for individualized and visualized interpretation. SHAP addressed the cognitive opacity of machine learning models.

Results

Eleven radiomics features were used to calculate the radiomics score. Five critical elements were used to build the XGBoost model: capsular invasion, radiomics score, diameter, age, and calcification. The area under the curve was 91.53% and 90.88% in the training and test cohorts, respectively. SHAP plots showed the influence of each parameter on the XGBoost model, including positive (i.e., capsular invasion, radiomics score, diameter, and calcification) and negative (i.e., age) impacts. The XGBoost model outperformed the radiologist, increasing the AUC by 44%.

Conclusions

The radiomics-based XGBoost model predicted CCLNM in PTC patients. Visual interpretation using SHAP made the model an effective tool for preoperative guidance of clinical procedures, including positive and negative impacts.

Identification of key predictors of hospital mortality in critically ill patients with embolic stroke using machine learning

Embolic stroke (ES) is characterized by high morbidity and mortality. Its mortality predictors remain unclear. The present study aimed to use machine learning (ML) to identify the key predictors of mortality for ES patients in the intensive care unit (ICU). Data were extracted from two large ICU databases: Medical Information Mart for Intensive Care (MIMIC)-IV for training and internal validation, and eICU Collaborative Research Database (eICU-CRD) for external validation. We developed predictive models of ES mortality based on 15 ML algorithms. We relied on the synthetic minority oversampling technique (SMOTE) to address class imbalance. Our main performance metric was area under the receiver operating characteristic (AUROC). We adopted recursive feature elimination (RFE) for feature selection. We assessed model performance using three disease-severity scoring systems as benchmarks. Of the 1566 and 207 ES patients enrolled in the two databases, there were 173 (15.70%), 73 (15.57%), and 36 (17.39%) hospital mortality in the training, internal validation, and external validation cohort, respectively. The random forest (RF) model had the largest AUROC (0.806) in the internal validation phase and was chosen as the best model. The AUROC of the RF compact (RF-COM) model containing the top six features identified by RFE was 0.795. In the external validation phase, the AUROC of the RF model was 0.838, and the RF-COM model was 0.830, outperforming other models. Our findings suggest that the RF model was the best model and the top six predictors of ES hospital mortality were Glasgow Coma Scale, white blood cell, blood urea nitrogen, bicarbonate, age, and mechanical ventilation.

Cuproptosis status affects treatment options about immunotherapy and targeted therapy for patients with kidney renal clear cell carcinoma

The development of immunotherapy has changed the treatment landscape of advanced kidney renal clear cell carcinoma (KIRC), offering patients more treatment options. Cuproptosis, a novel cell death mode dependent on copper ions and mitochondrial respiration has not yet been studied in KIRC. We assembled a comprehensive cohort of The Cancer Genome Atlas (TCGA)-KIRC and GSE29609, performed cluster analysis for typing twice using seven cuproptosis-promoting genes (CPGs) as a starting point, and assessed the differences in biological and clinicopathological characteristics between different subtypes. Furthermore, we explored the tumor immune infiltration landscape in KIRC using ESTIMATE and single-sample gene set enrichment analysis (ssGSEA) and the potential molecular mechanisms of cuproptosis in KIRC using enrichment analysis. We constructed a cuproptosis score (CUS) using the Boruta algorithm combined with principal component analysis. We evaluated the impact of CUS on prognosis, targeted therapy, and immunotherapy in patients with KIRC using survival analysis, the predictions from the Cancer Immunome Atlas database, and targeted drug susceptibility analysis. We found that patients with high CUS levels show poor prognosis and efficacy against all four immune checkpoint inhibitors, and their immunosuppression may depend on TGFB1. However, the high-CUS group showed higher sensitivity to sunitinib, axitinib, and elesclomol. Sunitinib monotherapy may reverse the poor prognosis and result in higher progression free survival. Then, we identified two potential CPGs and verified their differential expression between the KIRC and the normal samples. Finally, we explored the effect of the key gene FDX1 on the proliferation of KIRC cells and confirmed the presence of cuproptosis in KIRC cells. We developed a targeted therapy and immunotherapy strategy for advanced KIRC based on CUS. Our findings provide new insights into the relationship among cuproptosis, metabolism, and immunity in KIRC.

Subtyping and grading of lower-grade gliomas using integrated feature selection and support vector machine

Classifying lower-grade gliomas (LGGs) is a crucial step for accurate therapeutic intervention. The histopathological classification of various subtypes of LGG, including astrocytoma, oligodendroglioma and oligoastrocytoma, suffers from intraobserver and interobserver variability leading to inaccurate classification and greater risk to patient health. We designed an efficient machine learning-based classification framework to diagnose LGG subtypes and grades using transcriptome data. First, we developed an integrated feature selection method based on correlation and support vector machine (SVM) recursive feature elimination. Then, implementation of the SVM classifier achieved superior accuracy compared with other machine learning frameworks. Most importantly, we found that the accuracy of subtype classification is always high (>90%) in a specific grade rather than in mixed grade (~80%) cancer. Differential co-expression analysis revealed higher heterogeneity in mixed grade cancer, resulting in reduced prediction accuracy. Our findings suggest that it is necessary to identify cancer grades and subtypes to attain a higher classification accuracy. Our six-class classification model efficiently predicts the grades and subtypes with an average accuracy of 91% (±0.02). Furthermore, we identify several predictive biomarkers using co-expression, gene set enrichment and survival analysis, indicating our framework is biologically interpretable and can potentially support the clinician.

Prediction poverty levels of needy college students using RF-PCA model

Nowadays, poverty-stricken college students have become a special group among college students and occupied a higher proportion in it. How to accurately identify poverty levels of college students and provide funding is a new problem for universities. In this study, a novel model, which incorporated Random Forest with Principle Components Analysis (RF-PCA), is proposed to predict poverty levels of college students. To establish this model, we collect some useful information is to construct the datasets which include 4 classes of poverty levels and 21 features of poverty-stricken college students. Furthermore, the feature dimension reduction consists of two steps: the first step is to select the top 16 features with the ranking of feature, according to the Gini importance and Shapley Additive explanations (SHAP) values of features based on Random Forest (RF) model; the second step is to extract 11 dimensions by means of Principle Components Analysis (PCA). Subsequently, confusion metrics and receiver operating characteristic (ROC) curves are utilized to evaluate the promising performance of the proposed model. Especially the accuracy of the model achieves 78.61% . Finally, compared with seven states of the art classification algorithms, the proposed model achieves a higher prediction accuracy, which indicates that the results provide great potential to identify the poverty levels of college students.

An Interpretable Machine Learning Approach to Predict Fall Risk Among Community-Dwelling Older Adults: a Three-Year Longitudinal Study

BACKGROUND

Adverse health effects resulting from falls are a major public health concern. Although studies have identified risk factors for falls, none have examined long-term prediction of fall risk. Furthermore, recent evidence suggests that there are additional risk factors, such as psychosocial factors.

OBJECTIVE

In this 3-year longitudinal study, we evaluated a predictive model for risk of fall among community-dwelling older adults using machine learning methods.

DESIGN

A 3-year follow-up prospective longitudinal study (from 2010 to 2013).

SETTING

Twenty-four municipalities in nine of the 47 prefectures (provinces) of Japan.

PARTICIPANTS

Community-dwelling individuals aged ≥65 years who were functionally independent at baseline (n = 61,883).

METHODS

The baseline survey was conducted from August 2010 to January 2012, and the follow-up survey was conducted from October to December 2013. Both surveys were conducted involving self-reported questionnaires. The measured outcome at the follow-up survey was self-reported multiple falls during the previous year. The 142 variables included in the baseline survey were regarded as candidate predictors. The random-forest-based Boruta algorithm was used to select predictors, and the eXtreme Gradient Boosting algorithm with 10 repetitions of nested k-fold cross-validation was used for modeling and model evaluation. Furthermore, we used shapley additive explanations to gain insight into the behavior of the prediction model.

KEY RESULTS

Fourteen out of 142 candidate features were selected as predictors. Among these predictors, experience of falling as of the baseline survey was the most important feature, followed by self-rated health and age. Moreover, sense of coherence was newly identified as a risk factor for falls.

CONCLUSIONS

This study suggests that machine learning tools can be adapted to explore new associative factors, make accurate predictions, and provide actionable insights for fall prevention strategies.

Gut microbiota dysbiosis induced by polychlorinated biphenyl 126 contributes to increased brain proinflammatory cytokines: Landscapes from the gut-brain axis and fecal microbiota transplantation

The pathogenesis of brain inflammation induced by polychlorinated biphenyl 126 (PCB126) has not yet been fully illustrated. Growing evidence highlights the relevance of the microbiota-gut-brain axis in central nervous system (CNS) dysfunction. Therefore, we aimed to study the role of the gut microbiota in PCB126-induced proinflammatory cytokine increases in the mouse brain. The results showed that PCB126 exposure significantly disordered gut bacterial communities, resulting in the enrichment of gram-negative bacteria (e.g., Bacteroidetes and Proteobacteria), further leading to elevated levels of the gram-negative bacterial lipopolysaccharide (LPS). Subsequently, colonic toll-like receptor 4 (TLR-4) was activated by bacterial LPS, which promoted proinflammatory cytokine generation and inhibited tight junction (TJ) protein expression. Then, bacterial LPS translocated from the gut lumen into the blood circulation and reached the brain, triggering LPS/TLR-4-mediated increases in brain proinflammatory cytokines. Further analysis after fecal microbiota transplantation (FMT) suggested that the gut microbiota disturbance caused by PCB126 could induce elevated bacterial LPS and trigger TLR-4-mediated increases in proinflammatory cytokines in the brain. This study highlights the possibility that PCB126-induced gut microbiota disorder contributes to increased brain proinflammatory cytokines. These results provide a new perspective for identifying the toxicity mechanisms of PCB126 and open up the possibility of modulating the gut microbiota as a therapeutic target for CNS disease caused by environmental pollution.

A Novel Necroptosis-Related Prognostic Signature of Glioblastoma Based on Transcriptomics Analysis and Single Cell Sequencing Analysis

Background: Glioblastoma (GBM) is the most common and deadly brain tumor. The clinical significance of necroptosis (NCPS) genes in GBM is unclear. The goal of this study is to reveal the potential prognostic NCPS genes associated with GBM, elucidate their functions, and establish an effective prognostic model for GBM patients. Methods: Firstly, the NCPS genes in GBM were identified by single-cell analysis of the GSE182109 dataset in the GEO database and weighted co-expression network analysis (WGCNA) of The Cancer Genome Atlas (TCGA) data. Three machine learning algorithms (Lasso, SVM-RFE, Boruta) combined with COX regression were used to build prognostic models. The subsequent analysis included survival, immune microenvironments, and mutations. Finally, the clinical significance of NCPS in GBM was explored by constructing nomograms. Results: We constructed a GBM prognostic model composed of NCPS-related genes, including CTSD, AP1S1, YWHAG, and IER3, which were validated to have good performance. According to the above prognostic model, GBM patients in the TCGA and CGGA groups could be divided into two groups according to NCPS, with significant differences in survival analysis between the two groups and a markedly worse prognostic status in the high NCPS group (p < 0.001). In addition, the high NCPS group had higher levels of immune checkpoint-related gene expression, suggesting that they may be more likely to benefit from immunotherapy. Conclusions: Four genes (CTSD, AP1S1, YWHAG, and IER3) were screened through three machine learning algorithms to construct a prognostic model for GBM. These key and novel diagnostic markers may become new targets for diagnosing and treating patients with GBM.

An Improved Air Quality Index Machine Learning-Based Forecasting with Multivariate Data Imputation Approach

Accurate, timely air quality index (AQI) forecasting helps industries in selecting the most suitable air pollution control measures and the public in reducing harmful exposure to pollution. This article proposes a comprehensive method to forecast AQIs. Initially, the work focused on predicting hourly ambient concentrations of PM2.5 and PM10 using artificial neural networks. Once the method was developed, the work was extended to the prediction of other criteria pollutants, i.e., O3, SO2, NO2, and CO, which fed into the process of estimating AQI. The prediction of the AQI not only requires the selection of a robust forecasting model, it also heavily relies on a sequence of pre-processing steps to select predictors and handle different issues in data, including gaps. The presented method dealt with this by imputing missing entries using missForest, a machine learning-based imputation technique which employed the random forest (RF) algorithm. Unlike the usual practice of using RF at the final forecasting stage, we utilized RF at the data pre-processing stage, i.e., missing data imputation and feature selection, and we obtained promising results. The effectiveness of this imputation method was examined against a linear imputation method for the six criteria pollutants and the AQI. The proposed approach was validated against ambient air quality observations for Al-Jahra, a major city in Kuwait. Results obtained showed that models trained using missForest-imputed data could generalize AQI forecasting and with a prediction accuracy of 92.41% when tested on new unseen data, which is better than earlier findings.

Texture analysis in brain T2 and diffusion MRI differentiates histology-verified grey and white matter pathology types in multiple sclerosis

BACKGROUND

Multiple sclerosis (MS) is a co mplex disease of the central nervous system involving several types of brain pathology that are difficult to characterize using conventional imaging methods.

NEW METHOD

We originated novel texture analysis and machine learning approaches for classifying MS pathology subtypes as compared with 2 common advanced MRI measures: magnetization transfer ratio (MTR) and fractional anisotropy (FA). Texture analysis used an optimized grey level co-occurrence matrix method with histology-informed 7T T2-weighted magnetic resonance imaging (MRI) and diffusion tensor imaging (DTI) from 15 MS and 12 control brain specimens. DTI analysis took an innovative approach that assessed the texture across diffusion directions upsampled from 30 to 90. Tissue types included de- and re-myelinated lesions and normal-appearing areas in both grey and white matter, and diffusely abnormal white matter. Data analyses were stepwise, including: (1) group-wise classification using random forest algorithms based on all or individual imaging parameters; (2) parameter importance ranking; and (3) pairwise analysis using top-ranked features.

RESULTS

Texture analysis performed better than MTR and FA, with T2 texture performed the best. T2 texture measures ranked the highest in classifying most grey and white matter tissue types, including de- versus re-myelinated lesions and among grey matter lesion subtypes (accuracy=0.86-0.59; kappa=0.60-0.41). Diffusion texture best differentiated normal appearing and control white matter.

COMPARISON WITH EXISTING METHODS

There is no established method in imaging for differentiating MS pathology subtypes. In combined texture analysis and machine learning studies, there is also no direct evidence comparing conventional with advanced MRI measures for assessing MS pathology. Further, this study is unique in conducting innovative texture analysis with DTI following data-augmentation using robust methods.

CONCLUSIONS

T2 and diffusion MRI texture analysis integrated with machine learning may be valuable approaches for characterizing MS pathology.

CRCF: A Method of Identifying Secretory Proteins of Malaria Parasites

Malaria is a mosquito-borne disease that results in millions of cases and deaths annually. The development of a fast computational method that identifies secretory proteins of the malaria parasite is important for research on antimalarial drugs and vaccines. Thus, a method was developed to identify the secretory proteins of malaria parasites. In this method, a reduced alphabet was selected to recode the original protein sequence. A feature synthesis method was used to synthesise three different types of feature information. Finally, the random forest method was used as a classifier to identify the secretory proteins. In addition, a web server was developed to share the proposed algorithm. Experiments using the benchmark dataset demonstrated that the overall accuracy achieved by the proposed method was greater than 97.8 percent using the 10-fold cross-validation method. Furthermore, the reduced schemes and characteristic performance analyses are discussed.

Identifying adverse childhood experiences with electronic health records of linked mothers and children in England: a multistage development and validation study

BACKGROUND

Electronic health records (EHRs) of mothers and children provide an opportunity to identify adverse childhood experiences (ACEs) during crucial periods of childhood development, yet well developed indicators of ACEs remain scarce. We aimed to develop clinically relevant indicators of ACEs for linked EHRs of mothers and children using a multistage prediction model of child maltreatment and maternal intimate partner violence (IPV).

METHODS

In this multistage development and validation study, we developed a representative population-based birth cohort of mothers and children in England, followed from up to 2 years before birth to up to 5 years after birth across the Clinical Practice Research Datalink (CPRD) GOLD (primary care), Hospital Episode Statistics (secondary care), and the Office for National Statistics mortality register. We included livebirths in England between July 1, 2004, and June 30, 2016, to mothers aged 16-55 years, who had registered with a general practitioner (GP) that met CPRD quality standards before 21 weeks of gestation. The primary outcome (reference standard) was any child maltreatment or maternal IPV in either the mother's or child's record from 2 years before birth (maternal IPV only) to 5 years after birth. We used seven prediction models, combined with expert ratings, to systematically develop indicators. We validated the final indicators by integrating results from machine learning models, survival analyses, and clustering analyses in the validation cohort.

FINDINGS

We included data collected between July 1, 2002, and June 27, 2018. Of 376 006 eligible births, we included 211 393 mother-child pairs (422 786 patients) from 400 practices, of whom 126 837 mother-child pairs (60·0%; 240 practices) were randomly assigned to a derivation cohort and 84 556 pairs (40·0%; 160 practices) to a validation cohort. We included 63 indicators in six ACE domains: maternal mental health problems, maternal substance misuse, adverse family environments, child maltreatment, maternal IPV, and high-risk presentations of child maltreatment. Excluding the seven indicators in the reference standard, 56 indicators showed high discriminative validity for the reference standard of any child maltreatment or maternal IPV between 2 years before and 5 years after birth (validation cohort, area under the receiver operating characteristic curve 0·85 [95% CI 0·84-0·86]). During the 2 years before birth and 5 years after birth, the overall period prevalence of maternal IPV and child maltreatment (reference standard) was 2·3% (2876 of 126 837 pairs) in the derivation cohort and 2·3% (1916 of 84 556 pairs) in the validation cohort. During the 2 years before and after birth, the period prevalence was 39·1% (95% CI 38·7-39·5; 34 773 pairs) for any of the 63 ACE indicators, 22·2% (21·8-22·5%; 20 122 pairs) for maternal mental health problems, 15·7% (15·4-16·0%; 14 549 pairs) for adverse family environments, 8·1% (7·8-8·3%; 6808 pairs) for high-risk presentations of child maltreatment, 6·9% (6·7-7·2%; 7856 pairs) for maternal substance misuse, and 3·0% (2·9-3·2%; 2540 pairs) for any child maltreatment (2·4% [2·3-5·6%; 2051 pairs]) and maternal IPV (1·0% [0·8-1·0%; 875 pairs]). 62·6% (21 785 of 34 773 pairs) of ACEs were recorded in primary care only, and 72·3% (25 140 cases) were recorded in the maternal record only.

INTERPRETATION

We developed clinically relevant indicators for identifying ACEs using the EHRs of mothers and children presenting to general practices and hospital admissions. Over 70% of ACEs were identified via maternal records and were recorded in primary care by GPs within 2 years of birth, reinforcing the importance of reviewing parental and carer records to inform clinical responses to children. ACE indicators can contribute to longitudinal surveillance informing public health policy and resource allocation. Further evaluation is required to determine how ACE indicators can be used in clinical practice.

FUNDING

None.

Machine learning-based classification of the movements of children with profound or severe intellectual or multiple disabilities using environment data features

Communication interventions have broadened from dialogical meaning-making, assessment approaches, to remote-controlled interactive objects. Yet, interpretation of the mostly pre-or protosymbolic, distinctive, and idiosyncratic movements of children with intellectual disabilities (IDs) or profound intellectual and multiple disabilities (PIMD) using computer-based assistive technology (AT), machine learning (ML), and environment data (ED: location, weather indices and time) remain insufficiently unexplored. We introduce a novel behavior inference computer-based communication-aid AT system structured on machine learning (ML) framework to interpret the movements of children with PIMD/IDs using ED. To establish a stable system, our study aimed to train, cross-validate (10-fold), test and compare the classification accuracy performance of ML classifiers (eXtreme gradient boosting [XGB], support vector machine [SVM], random forest [RF], and neural network [NN]) on classifying the 676 movements to 2, 3, or 7 behavior outcome classes using our proposed dataset recalibration (adding ED to movement datasets) with or without Boruta feature selection (53 child characteristics and movements, and ED-related features). Natural-child-caregiver-dyadic interactions observed in 105 single-dyad video-recorded (30-hour) sessions targeted caregiver-interpreted facial, body, and limb movements of 20 8-to 16-year-old children with PIMD/IDs and simultaneously app-and-sensor-collected ED. Classification accuracy variances and the influences of and the interaction among recalibrated dataset, feature selection, classifiers, and classes on the pooled classification accuracy rates were evaluated using three-way ANOVA. Results revealed that Boruta and NN-trained dataset in class 2 and the non-Boruta SVM-trained dataset in class 3 had >76% accuracy rates. Statistically significant effects indicating high classification rates (>60%) were found among movement datasets: with ED, non-Boruta, class 3, SVM, RF, and NN. Similar trends (>69%) were found in class 2, NN, Boruta-trained movement dataset with ED, and SVM and RF, and non-Boruta-trained movement dataset with ED in class 3. These results support our hypotheses that adding environment data to movement datasets, selecting important features using Boruta, using NN, SVM and RF classifiers, and classifying movements to 2 and 3 behavior outcomes can provide >73.3% accuracy rates, a promising performance for a stable ML-based behavior inference communication-aid AT system for children with PIMD/IDs.

Anosmia Testing as Early Detection of SARS-CoV-2 Positivity; A Prospective Study under Screening Conditions

Sudden onset of anosmia is a phenomenon often associated with developing COVID-19 disease and has even been described as an initial isolated symptom in individual cases. In this case-control study, we investigated the feasibility of this condition as a suitable screening test in a population at risk. We performed a prospective study with a total of 313 subjects with suspected SARS-CoV-2 infection. In parallel to routine PCR analysis, a modified commercial scent test was performed to objectify the presence of potential anosmia as a predictor of SARS-CoV-2 positivity. Furthermore, a structured interview assessment of the participants was conducted. A total of 12.1% of the study participants had molecular genetic detection of SARS-CoV-2 infection in the nasopharyngeal swab. It could be demonstrated that these subjects had a significantly weaker olfactory identification performance of the scents. Further analysis of the collected data from the scent test and medical history via random forest (Boruta) algorithm showed that no improvement of the prediction power was achieved by this design. The assay investigated in this study may be suitable for screening general olfactory function. For the screening of COVID-19, it seems to be affected by too many external and internal biases and requires too elaborate and selective pre-test screening.

Vascular Dysfunction Is Central to Alzheimer's Disease Pathogenesis in APOE e4 Carriers

Alzheimer's disease (AD) is the most common form of dementia and the leading risk factor, after age, is possession of the apolipoprotein E epsilon 4 allele (APOE4). Approximately 50% of AD patients carry one or two copies of APOE4 but the mechanisms by which it confers risk are still unknown. APOE4 carriers are reported to demonstrate changes in brain structure, cognition, and neuropathology, but findings have been inconsistent across studies. In the present study, we used multi-modal data to characterise the effects of APOE4 on the brain, to investigate whether AD pathology manifests differently in APOE4 carriers, and to determine if AD pathomechanisms are different between carriers and non-carriers. Brain structural differences in APOE4 carriers were characterised by applying machine learning to over 2000 brain MRI measurements from 33,384 non-demented UK biobank study participants. APOE4 carriers showed brain changes consistent with vascular dysfunction, such as reduced white matter integrity in posterior brain regions. The relationship between APOE4 and AD pathology was explored among the 1260 individuals from the Religious Orders Study and Memory and Aging Project (ROSMAP). APOE4 status had a greater effect on amyloid than tau load, particularly amyloid in the posterior cortical regions. APOE status was also highly correlated with cerebral amyloid angiopathy (CAA). Bulk tissue brain transcriptomic data from ROSMAP and a similar dataset from the Mount Sinai Brain Bank showed that differentially expressed genes between the dementia and non-dementia groups were enriched for vascular-related processes (e.g., "angiogenesis") in APOE4 carriers only. Immune-related transcripts were more strongly correlated with AD pathology in APOE4 carriers with some transcripts such as TREM2 and positively correlated with pathology severity in APOE4 carriers, but negatively in non-carriers. Overall, cumulative evidence from the largest neuroimaging, pathology, and transcriptomic studies available suggests that vascular dysfunction is key to the development of AD in APOE4 carriers. However, further studies are required to tease out non-APOE4-specific mechanisms.

Screening of Gene Expression Markers for Corona Virus Disease 2019 Through Boruta_MCFS Feature Selection

Since the first report of SARS-CoV-2 virus in Wuhan, China in December 2019, a global outbreak of Corona Virus Disease 2019 (COVID-19) pandemic has been aroused. In the prevention of this disease, accurate diagnosis of COVID-19 is the center of the problem. However, due to the limitation of detection technology, the test results are impossible to be totally free from pseudo-positive or -negative. Improving the precision of the test results asks for the identification of more biomarkers for COVID-19. On the basis of the expression data of COVID-19 positive and negative samples, we first screened the feature genes through ReliefF, minimal-redundancy-maximum-relevancy, and Boruta_MCFS methods. Thereafter, 36 optimal feature genes were selected through incremental feature selection method based on the random forest classifier, and the enriched biological functions and signaling pathways were revealed by Gene Ontology and Kyoto Encyclopedia of Genes and Genomes. Also, protein-protein interaction network analysis was performed on these feature genes, and the enriched biological functions and signaling pathways of main submodules were analyzed. In addition, whether these 36 feature genes could effectively distinguish positive samples from the negative ones was verified by dimensionality reduction analysis. According to the results, we inferred that the 36 feature genes selected via Boruta_MCFS could be deemed as biomarkers in COVID-19.

Using random forest to identify longitudinal predictors of health in a 30-year cohort study

Due to the wealth of exposome data from longitudinal cohort studies that is currently available, the need for methods to adequately analyze these data is growing. We propose an approach in which machine learning is used to identify longitudinal exposome-related predictors of health, and illustrate its potential through an application. Our application involves studying the relation between exposome and self-perceived health based on the 30-year running Doetinchem Cohort Study. Random Forest (RF) was used to identify the strongest predictors due to its favorable prediction performance in prior research. The relation between predictors and outcome was visualized with partial dependence and accumulated local effects plots. To facilitate interpretation, exposures were summarized by expressing them as the average exposure and average trend over time. The RF model's ability to discriminate poor from good self-perceived health was acceptable (Area-Under-the-Curve = 0.707). Nine exposures from different exposome-related domains were largely responsible for the model's performance, while 87 exposures seemed to contribute little to the performance. Our approach demonstrates that ML can be interpreted more than widely believed, and can be applied to identify important longitudinal predictors of health over the life course in studies with repeated measures of exposure. The approach is context-independent and broadly applicable.

Effects of Driving Factors on Forest Aboveground Biomass (AGB) in China’s Loess Plateau by Using Spatial Regression Models

Forests are the main body of carbon sequestration in terrestrial ecosystems and forest aboveground biomass (AGB) is an important manifestation of forest carbon sequestration. Reasonable and accurate quantification of the relationship between AGB and its driving factors is of great importance for increasing the biomass and function of forests. Remote sensing observations and field measurements can be used to estimate AGB in large areas. To explore the applicability of the panel data models in AGB and its driving factors, we compared the results of panel data models (spatial error model and spatial lag model) with those of geographically weighted regression (GWR) and ordinary least squares (OLS) to quantify the relationship between AGB and its driving factors. Furthermore, we estimated the tree height, diameter at breast height, canopy cover (CC) and species diversity index (Shannon–Wiener index) of Robinia pseudoacacia plantations in Changwu on the Loess Plateau using field data and remote sensing images by a random forest model and estimated soil organic carbon (SOC) contents using laboratory data by ordinary kriging (OK) interpolation. We estimated AGB using the already estimated tree height and diameter at breast height combined with the allometric growth equation. In this study, we estimated SOC contents by OK interpolation, and the accuracy R2 values for each soil layer were greater than 0.81. We estimated diameter at breast height (DBH), CC, SW and tree height (TH) using the random forest, and the accuracy R2 values were 0.85, 0.82, 0.76 and 0.68, respectively. We estimated AGB with random forest and the allometric growth equation and found that the average AGB was 55.80 t/ha. The OLS results showed that the residuals of the OLS regression exhibited obvious spatial correlations and rejected OLS applications. GWR, SEM and SLM were used for spatial regression analysis, and SEM was the best model for explaining the relationship between AGB and its driving factors. We also found that AGB was significantly positively correlated with CC, SW, and 0–60 cm SOC content (p < 0.05) and significantly negatively correlated with slope aspect (p < 0.01). This study provides a new idea for studying the relationship between AGB and its driving factors and provides a basis for practical forest management, increasing biomass, and giving full play to the role of carbon sequestration.

Integrated Bioinformatics Identifies FREM1 as a Diagnostic Gene Signature for Heart Failure

Objective

This study is aimed at integrating bioinformatics and machine learning to determine novel diagnostic gene signals in the progression of heart failure disease.

Methods

The heart failure microarray datasets and RNA-seq datasets have been downloaded from the public database. Differentially expressed genes (DE genes) are screened out, and then, we analyze their biological functions and pathways. Integrating three machine learning methods, the least absolute shrinkage and selection operator (LASSO) algorithm, random forest (RF) algorithm, and support vector machine recursive feature elimination (SVM-RFE) are used to determine candidate diagnostic gene signals. Then, external independent RNA-seq datasets evaluate the diagnostic value of gene signals. Finally, the convolution tool CIBERSORT estimated the composition pattern of immune cell subtypes in heart failure and carried out a correlation analysis combined with gene signals.

Results

Under the set threshold, we obtained 47 DE genes with the most significant differences. Enrichment analysis shows that most of them are related to hypertrophy, matrix structural constituent, protein binding, inflammatory immune pathway, cardiovascular disease, and inflammatory disease. Three machine learning methods assisted in determining the potential characteristic signals Fras1-related extracellular matrix 1 (FREM1) and meiosis-specific nuclear structural 1 (MNS1). Validation of external datasets confirms that FREM1 is a diagnostic gene signal for heart failure. Immune cell subtypes of tissue specimens found T cell CD8, mast cell resting, T cell CD4 memory resting, T cell regulation (Tregs), monocytes, macrophages M2, T cell CD4 naive, macrophages M0, and neutrophils are associated with HF.

Conclusion

The gene signal FREM1 may be a potential molecular target in the development of HF and is related to the difference in immune infiltration of HF tissue.

Impact of Freeze-Drying on the Determination of the Geographical Origin of Almonds (Prunus dulcis Mill.) by Near-Infrared (NIR) Spectroscopy

Near-infrared (NIR) spectroscopy is a proven tool for the determination of food authenticity, mainly because of good classification results and the possibility of industrial use due to its easy and fast application. Since water shows broad absorption bands, the water content of a sample should be as low as possible. Freeze-drying is a commonly used preparatory step for this to reduce the water content in the sample. However, freeze-drying, also known as lyophilization, is very time-consuming impeding the widespread usage of NIR analysis as a rapid method for incoming goods inspections. We used a sample set of 72 almond samples from six economically relevant almond-producing countries to investigate the question of how important lyophilization is to obtain a well-performing classification model. For this approach, the samples were ground and lyophilized for 3 h, 24 h, and 48 h and compared to non-freeze-dried samples. Karl-Fischer titration of non-lyophilized samples showed that water contents ranged from 3.0 to 10.5% and remained constant at 0.36 ± 0.13% after a freeze-drying period of 24 h. The non-freeze-dried samples showed a classification accuracy of 93.9 ± 6.4%, which was in the same range as the samples which were freeze-dried for 3 h (94.2 ± 7.8%), 24 h (92.5 ± 8.7%), and 48 h (95.0 ± 9.0%). Feature selection was performed using the Boruta algorithm, which showed that signals from lipids and proteins are relevant for the origin determination. The presented study showed that samples with low water content, especially nuts, can be analyzed without the time-consuming preparation step of freeze-drying to obtain robust and fast results, which are especially required for incoming goods inspection.

Improving the Stability of the Variable Selection with Small Datasets in Classification and Regression Tasks

Within the design of a machine learning-based solution for classification or regression problems, variable selection techniques are often applied to identify the input variables, which mainly affect the considered target. The selection of such variables provides very interesting advantages, such as lower complexity of the model and of the learning algorithm, reduction of computational time and improvement of performances. Moreover, variable selection is useful to gain a profound knowledge of the considered problem. High correlation in variables often produces multiple subsets of equally optimal variables, which makes the traditional method of variable selection unstable, leading to instability and reducing the confidence of selected variables. Stability identifies the reproducibility power of the variable selection method. Therefore, having a high stability is as important as the high precision of the developed model. The paper presents an automatic procedure for variable selection in classification (binary and multi-class) and regression tasks, which provides an optimal stability index without requiring any a priori information on data. The proposed approach has been tested on different small datasets, which are unstable by nature, and has achieved satisfactory results.

Network-based cancer heterogeneity analysis incorporating multi-view of prior information

MOTIVATION

Cancer genetic heterogeneity analysis has critical implications for tumour classification, response to therapy and choice of biomarkers to guide personalized cancer medicine. However, existing heterogeneity analysis based solely on molecular profiling data usually suffers from a lack of information and has limited effectiveness. Many biomedical and life sciences databases have accumulated a substantial volume of meaningful biological information. They can provide additional information beyond molecular profiling data, yet pose challenges arising from potential noise and uncertainty.

RESULTS

In this study, we aim to develop a more effective heterogeneity analysis method with the help of prior information. A network-based penalization technique is proposed to innovatively incorporate a multi-view of prior information from multiple databases, which accommodates heterogeneity attributed to both differential genes and gene relationships. To account for the fact that the prior information might not be fully credible, we propose a weighted strategy, where the weight is determined dependent on the data and can ensure that the present model is not excessively disturbed by incorrect information. Simulation and analysis of The Cancer Genome Atlas glioblastoma multiforme data demonstrate the practical applicability of the proposed method.

AVAILABILITY AND IMPLEMENTATION

R code implementing the proposed method is available at https://github.com/mengyunwu2020/PECM. The data that support the findings in this paper are openly available in TCGA (The Cancer Genome Atlas) at https://portal.gdc.cancer.gov/.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Microbiota alterations in proline metabolism impact depression

The microbiota-gut-brain axis has emerged as a novel target in depression, a disorder with low treatment efficacy. However, the field is dominated by underpowered studies focusing on major depression not addressing microbiome functionality, compositional nature, or confounding factors. We applied a multi-omics approach combining pre-clinical models with three human cohorts including patients with mild depression. Microbial functions and metabolites converging onto glutamate/GABA metabolism, particularly proline, were linked to depression. High proline consumption was the dietary factor with the strongest impact on depression. Whole-brain dynamics revealed rich club network disruptions associated with depression and circulating proline. Proline supplementation in mice exacerbated depression along with microbial translocation. Human microbiota transplantation induced an emotionally impaired phenotype in mice and alterations in GABA-, proline-, and extracellular matrix-related prefrontal cortex genes. RNAi-mediated knockdown of proline and GABA transporters in Drosophila and mono-association with L. plantarum, a high GABA producer, conferred protection against depression-like states. Targeting the microbiome and dietary proline may open new windows for efficient depression treatment.

Metabolomics analysis reveals molecular linkages for the impact of vitamin D on childhood allergic airway diseases

BACKGROUND

Several studies have reported the relevance between serum vitamin D and allergic immunoglobulin E (IgE) responses and atopic diseases. However, a metabolomics-based approach to the impacts of vitamin D on allergic reactions remains unclear.

METHODS

A total of 111 children completed a 3-year follow-up were enrolled and classified based on longitudinal vitamin D status (≥ 30 ng/ml, n = 54; 20-29.9 ng/ml, n = 41; <20 ng/ml, n = 16). Urinary metabolomic profiling was performed using ¹ H-Nuclear magnetic resonance (NMR) spectroscopy at age 3. Integrative analyses of their associations related to vitamin D levels, atopic indices, and allergies were performed, and their roles in functional metabolic pathways were also assessed.

RESULTS

Six and five metabolites were identified to be significantly associated with vitamin D status and atopic diseases, respectively (FDR-adjusted p-value <.05). A further correlation analysis revealed that vitamin D-associated 3-hydroxyisobutyric acid and glutamine were positively correlated with atopic disease-associated succinic acid and alanine, respectively. Furthermore, hippuric acid was negatively correlated with atopic disease-associated formic acid, which was positively correlated with vitamin D level (p < .01). Absolute eosinophil count (AEC) was positively correlated with serum D. pteronyssinus- and D. farinae-specific IgE level (p < .01) but negatively correlated with vitamin D level (p < .05). Amino acid metabolisms were significantly associated with vitamin D related to childhood allergies.

CONCLUSION

Integrative metabolomic analysis provides the link of vitamin D-associated metabolites with the gut microbiome and immunoallergic reactions related to childhood allergies.

Novel Ensemble Feature Selection Approach and Application in Repertoire Sequencing Data

The T and B cell repertoire make up the adaptive immune system and is mainly generated through somatic V(D)J gene recombination. Thus, the VJ gene usage may be a potential prognostic or predictive biomarker. However, analysis of the adaptive immune system is challenging due to the heterogeneity of the clonotypes that make up the repertoire. To address the heterogeneity of the T and B cell repertoire, we proposed a novel ensemble feature selection approach and customized statistical learning algorithm focusing on the VJ gene usage. We applied the proposed approach to T cell receptor sequences from recovered COVID-19 patients and healthy donors, as well as a group of lung cancer patients who received immunotherapy. Our approach identified distinct VJ genes used in the COVID-19 recovered patients comparing to the healthy donors and the VJ genes associated with the clinical response in the lung cancer patients. Simulation studies show that the ensemble feature selection approach outperformed other state-of-the-art feature selection methods based on both efficiency and accuracy. It consistently yielded higher stability and sensitivity with lower false discovery rates. When integrated with different classification methods, the ensemble feature selection approach had the best prediction accuracy. In conclusion, the proposed novel approach and the integration procedure is an effective feature selection technique to aid in correctly classifying different subtypes to better understand the signatures in the adaptive immune response associated with disease or the treatment in order to improve treatment strategies.

Multisite and Multitemporal Grassland Yield Estimation Using UAV-Borne Hyperspectral Data

Grassland ecosystems can be hotspots of biodiversity and act as carbon sinks while at the same time providing the basis of forage production for ruminants in dairy and meat production. Annual grassland dry matter yield (DMY) is one of the most important agronomic parameters reflecting differences in usage intensity such as number of harvests and fertilization. Current methods for grassland DMY estimation are labor-intensive and prone to error due to small sample size. With the advent of unmanned aerial vehicles (UAVs) and miniaturized hyperspectral sensors, a novel tool for remote sensing of grassland with high spatial, temporal and radiometric resolution and coverage is available. The present study aimed at developing a robust model capable of estimating grassland biomass across a gradient of usage intensity throughout one growing season. Therefore, UAV-borne hyperspectral data from eight grassland sites in North Hesse, Germany, originating from different harvests, were utilized for the modeling of fresh matter yield (FMY) and DMY. Four machine learning (ML) algorithms were compared for their modeling performance. Among them, the rule-based ML method Cubist regression (CBR) performed best, delivering high prediction accuracies for both FMY (nRMSEp 7.6%, Rp2 0.87) and DMY (nRMSEp 12.9%, Rp2 0.75). The model showed a high robustness across sites and harvest dates. The best models were employed to produce maps for FMY and DMY, enabling the detailed analysis of spatial patterns. Although the complexity of the approach still restricts its practical application in agricultural management, the current study proved that biomass of grassland sites being subject to different management intensities can be modeled from UAV-borne hyperspectral data at high spatial resolution with high prediction accuracies.

Comprehensive Analysis of 5-Methylcytosine (m5C) Regulators and the Immune Microenvironment in Pancreatic Adenocarcinoma to Aid Immunotherapy

Background

Pancreatic adenocarcinoma (PAAD) is one of the most malignant cancers and has a poor prognosis. As a critical RNA modification, 5-methylcytosine (m⁵C) has been reported to regulate tumor progression, including PAAD progression. However, a comprehensive analysis of m⁵C regulators in PAAD is lacking.

Methods

In the present study, PAAD datasets were obtained from the Gene Expression Omnibus (GEO), The Cancer Genome Atlas (TCGA), International Cancer Genome Consortium (ICGC), and ArrayExpress databases. The expression pattern of m⁵C regulators were analyzed and patients were divided into different m⁵C clusters according to consensus clustering based on m⁵C regulators. Additionally, m⁵C differentially expressed genes (DEGs) were determined using Limma package. Based on m⁵C DEGs, patients were divided into m⁵C gene clusters. Moreover, m⁵C gene signatures were derived from m⁵C DEGs and a quantitative indicator, the m⁵C score, was developed from the m⁵C gene signatures.

Results

Our study showed that m⁵C regulators were differentially expressed in patients with PAAD. The m⁵C clusters and gene clusters based on m⁵C regulators and m⁵C DEGs were related to immune cell infiltration, immune-related genes and patient survival status, indicating that m⁵C modification play a central role in regulating PAAD development partly by modulating immune microenvironment. Additionally, a quantitative indicator, the m⁵C score, was also developed and was related to a series of immune-related indicators. Moreover, the m⁵C score precisely predicted the immunotherapy response and prognosis of patients with PAAD.

Conclusion

In summary, we confirmed that m⁵C regulators regulate PAAD development by modulating the immune microenvironment. In addition, a quantitative indicator, the m⁵C score, was developed to predict immunotherapy response and prognosis and assisted in identifying PAAD patients suitable for tailored immunotherapy strategies.

Decision tree–based machine learning analysis of intraoperative vasopressor use to optimize neurological improvement in acute spinal cord injury

OBJECTIVE

Previous work has shown that maintaining mean arterial pressures (MAPs) between 76 and 104 mm Hg intraoperatively is associated with improved neurological function at discharge in patients with acute spinal cord injury (SCI). However, whether temporary fluctuations in MAPs outside of this range can be tolerated without impairment of recovery is unknown. This retrospective study builds on previous work by implementing machine learning to derive clinically actionable thresholds for intraoperative MAP management guided by neurological outcomes.

METHODS

Seventy-four surgically treated patients were retrospectively analyzed as part of a longitudinal study assessing outcomes following SCI. Each patient underwent intraoperative hemodynamic monitoring with recordings at 5-minute intervals for a cumulative 28,594 minutes, resulting in 5718 unique data points for each parameter. The type of vasopressor used, dose, drug-related complications, average intraoperative MAP, and time spent in an extreme MAP range (< 76 mm Hg or > 104 mm Hg) were collected. Outcomes were evaluated by measuring the change in American Spinal Injury Association Impairment Scale (AIS) grade over the course of acute hospitalization. Features most predictive of an improvement in AIS grade were determined statistically by generating random forests with 10,000 iterations. Recursive partitioning was used to establish clinically intuitive thresholds for the top features.

RESULTS

At discharge, a significant improvement in AIS grade was noted by an average of 0.71 levels (p = 0.002). The hemodynamic parameters most important in predicting improvement were the amount of time intraoperative MAPs were in extreme ranges and the average intraoperative MAP. Patients with average intraoperative MAPs between 80 and 96 mm Hg throughout surgery had improved AIS grades at discharge. All patients with average intraoperative MAP > 96.3 mm Hg had no improvement. A threshold of 93 minutes spent in an extreme MAP range was identified after which the chance of neurological improvement significantly declined. Finally, the use of dopamine as compared to norepinephrine was associated with higher rates of significant cardiovascular complications (50% vs 25%, p < 0.001).

CONCLUSIONS

An average intraoperative MAP value between 80 and 96 mm Hg was associated with improved outcome, corroborating previous results and supporting the clinical verifiability of the model. Additionally, an accumulated time of 93 minutes or longer outside of the MAP range of 76–104 mm Hg is associated with worse neurological function at discharge among patients undergoing emergency surgical intervention for acute SCI.

Contribution of soil algae to the global carbon cycle

Soil photoautotrophic prokaryotes and micro-eukaryotes - known as soil algae - are, together with heterotrophic microorganisms, a constitutive part of the microbiome in surface soils. Similar to plants, they fix atmospheric carbon (C) through photosynthesis for their own growth, yet their contribution to global and regional biogeochemical C cycling still remains quantitatively elusive. Here, we compiled an extensive dataset on soil algae to generate a better understanding of their distribution across biomes and predict their productivity at a global scale by means of machine learning modelling. We found that, on average, (5.5 ± 3.4) × 10⁶ algae inhabit each gram of surface soil. Soil algal abundance especially peaked in acidic, moist and vegetated soils. We estimate that, globally, soil algae take up around 3.6 Pg C per year, which corresponds to c. 6% of the net primary production of terrestrial vegetation. We demonstrate that the C fixed by soil algae is crucial to the global C cycle and should be integrated into land-based efforts to mitigate C emissions.

Typical tumor immune microenvironment status determine prognosis in lung adenocarcinoma

•

Immune cells infiltration level in lung adenocarcinoma immune microenvironment were quantified and compared.

•

Three distinct tumor immune microenvironment subtypes were consistent with cancer immunity cycle in cancer dynamic development.

•

Immune infiltration status of three subtypes were correlated with significant mutated genes, copy number variation and cancer stemness

•

Prognostic biomarker lung adenocarcinoma immune microenvironment score model was constructed to assess immune infiltration status, evaluate immunotherapy response, and predict patient prognosis.

Background

Immune cells, vital components of tumor microenvironment, regulate tumor survival and progression. Lung adenocarcinoma (LUAD), the tumor with the highest mortality rate worldwide, reconstitutes tumor immune microenvironment (TIME) to avoid immune destruction. Data have shown that TIME influences LUAD prognosis and predicts immunotherapeutic efficacy. The related information about the role of TIME's characteristics in LUAD is limited.

Methods

We performed unsupervised consensus clustering via machine-learning techniques to identify TIME clusters among 1906 patients and gathered survival data. The characteristics of TIME clusters of LUAD were visualized by multi-omics analysis, pseudo-time dynamic analysis, and enrichment analysis. TIME score model was constructed by principal component analysis. Comprehensive analysis and validation were conducted to test the prognostic efficacy and immunotherapeutic response of TIME score.

Results

TIME clusters (A, B and C) were constructed and exhibited different immune infiltration states. Multi-omics analyses included significant mutated genes (SMG), copy number variation (CNV) and cancer stemness that were significantly different among the three clusters. TIME cluster A had a lower SMG, lower CNV, and lower stemness but a higher immune infiltration level compared to TIME clusters B and C. TIME score showed that patients in low TIME score group had higher overall survival rates, higher immune infiltration level and high expression of immune checkpoints. In validation cohorts, low TIME score subgroup had better drug sensitivity and favorable immunotherapeutic response.

Conclusion

We constructed a stable model of LUAD immune microenvironment characteristics that may improve the prognostic accuracy of patients, provide improved explanations of LUAD responses to immunotherapy, and provide new strategies for LUAD treatment.

Remote Sensing and Meteorological Data Fusion in Predicting Bushfire Severity: A Case Study from Victoria, Australia

The extent and severity of bushfires in a landscape are largely governed by meteorological conditions. An accurate understanding of the interactions of meteorological variables and fire behaviour in the landscape is very complex, yet possible. In exploring such understanding, we used 2693 high-confidence active fire points recorded by a Moderate Resolution Imaging Spectroradiometer (MODIS) sensor for nine different bushfires that occurred in Victoria between 1 January 2009 and 31 March 2009. These fires include the Black Saturday Bushfires of 7 February 2009, one of the worst bushfires in Australian history. For each fire point, 62 different meteorological parameters of bushfire time were extracted from Bureau of Meteorology Atmospheric high-resolution Regional Reanalysis for Australia (BARRA) data. These remote sensing and meteorological datasets were fused and further processed in assessing their relative importance using four different tree-based ensemble machine learning models, namely, Random Forest (RF), Fuzzy Forest (FF), Boosted Regression Tree (BRT), and Extreme Gradient Boosting (XGBoost). Google Earth Engine (GEE) and Landsat images were used in deriving the response variable–Relative Difference Normalised Burn Ratio (RdNBR), which was selected by comparing its performance against Difference Normalised Burn Ratio (dNBR). Our findings demonstrate that the FF algorithm utilising the Weighted Gene Coexpression Network Analysis (WGCNA) method has the best predictive performance of 96.50%, assessed against 10-fold cross-validation. The result shows that the relative influence of the variables on bushfire severity is in the following order: (1) soil moisture, (2) soil temperature, (3) air pressure, (4) air temperature, (5) vertical wind, and (6) relative humidity. This highlights the importance of soil meteorology in bushfire severity analysis, often excluded in bushfire severity research. Further, this study provides a scientific basis for choosing a subset of meteorological variables for bushfire severity prediction depending on their relative importance. The optimal subset of high-ranked variables is extremely useful in constructing simplified and computationally efficient surrogate models, which can be particularly useful for the rapid assessment of bushfire severity for operational bushfire management and effective mitigation efforts.

Integrated Machine Learning and Bioinformatic Analyses Constructed a Novel Stemness-Related Classifier to Predict Prognosis and Immunotherapy Responses for Hepatocellular Carcinoma Patients

Immunotherapy has made great progress in hepatocellular carcinoma (HCC), yet there is still a lack of biomarkers for predicting response to it. Cancer stem cells (CSCs) are the primary cause of the tumorigenesis, metastasis, and multi-drug resistance of HCC. This study aimed to propose a novel CSCs-related cluster of HCC to predict patients' response to immunotherapy. Based on RNA-seq datasets from The Cancer Genome Atlas (TCGA) and Progenitor Cell Biology Consortium (PCBC), one-class logistic regression (OCLR) algorithm was applied to compute the stemness index (mRNAsi) of HCC patients. Unsupervised consensus clustering was performed to categorize HCC patients into two stemness subtypes which further proved to be a predictor of tumor immune microenvironment (TIME) status, immunogenomic expressions and sensitivity to neoadjuvant therapies. Finally, four machine learning algorithms (LASSO, RF, SVM-RFE and XGboost) were applied to distinguish different stemness subtypes. Thus, a five-hub-gene based classifier was constructed in TCGA and ICGC HCC datasets to predict patients' stemness subtype in a more convenient and applicable way, and this novel stemness-based classification system could facilitate the prognostic prediction and guide clinical strategies of immunotherapy and targeted therapy in HCC.

Apparent Diffusion Coefficient Map-Based Texture Analysis for the Differentiation of Chromophobe Renal Cell Carcinoma from Renal Oncocytoma

Preoperative imaging differentiation between ChRCC and RO is difficult with conventional subjective evaluation, and the development of quantitative analysis is a clinical challenge. Forty-nine patients underwent partial or radical nephrectomy preceded by MRI and followed by pathological diagnosis with ChRCC or RO (ChRCC: n = 41, RO: n = 8). The whole-lesion volume of interest was set on apparent diffusion coefficient (ADC) maps of 1.5T-MRI. The importance of selected texture features (TFs) was evaluated, and diagnostic models were created using random forest (RF) analysis. The Mean Decrease Gini as calculated through RF analysis was the highest for mean_ADC_value. ChRCC had a significantly lower mean_ADC_value than RO (1.26 vs. 1.79 × 10−3 mm2/s, p < 0.0001). Feature selection by the Boruta method identified the first-quartile ADC value and GLZLM_HGZE as important features. ROC curve analysis showed that there was no significant difference in the classification performances between the mean_ADC_value-only model and the Boruta model (AUC: 0.954 vs. 0.969, p = 0.236). The mean ADC value had good predictive ability for the distinction between ChRCC and RO, comparable to that of the combination of TFs optimized for the evaluated cohort. The mean ADC value may be useful in distinguishing between ChRCC and RO.

Evaluation of tree-based statistical learning methods for constructing genetic risk scores

Background

Genetic risk scores (GRS) summarize genetic features such as single nucleotide polymorphisms (SNPs) in a single statistic with respect to a given trait. So far, GRS are typically built using generalized linear models or regularized extensions. However, these linear methods are usually not able to incorporate gene-gene interactions or non-linear SNP-response relationships. Tree-based statistical learning methods such as random forests and logic regression may be an alternative to such regularized-regression-based methods and are investigated in this article. Moreover, we consider modifications of random forests and logic regression for the construction of GRS.

Results

In an extensive simulation study and an application to a real data set from a German cohort study, we show that both tree-based approaches can outperform elastic net when constructing GRS for binary traits. Especially a modification of logic regression called logic bagging could induce comparatively high predictive power as measured by the area under the curve and the statistical power. Even when considering no epistatic interaction effects but only marginal genetic effects, the regularized regression method lead in most cases to inferior results.

Conclusions

When constructing GRS, we recommend taking random forests and logic bagging into account, in particular, if it can be assumed that possibly unknown epistasis between SNPs is present. To develop the best possible prediction models, extensive joint hyperparameter optimizations should be conducted.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12859-022-04634-w.