Clustering of samples and variables with mixed-type data

Topological Structures in the Space of Treatment-Naïve Patients with Chronic Lymphocytic Leukemia

Patients are complex and heterogeneous; clinical data sets are complicated by noise, missing data, and the presence of mixed-type data. Using such data sets requires understanding the high-dimensional "space of patients", composed of all measurements that define all relevant phenotypes. The current state-of-the-art merely defines spatial groupings of patients using cluster analyses. Our goal is to apply topological data analysis (TDA), a new unsupervised technique, to obtain a more complete understanding of patient space. We applied TDA to a space of 266 previously untreated patients with Chronic Lymphocytic Leukemia (CLL), using the "daisy" metric to compute distances between clinical records. We found clear evidence for both loops and voids in the CLL data. To interpret these structures, we developed novel computational and graphical methods. The most persistent loop and the most persistent void can be explained using three dichotomized, prognostically important factors in CLL: IGHV somatic mutation status, beta-2 microglobulin, and Rai stage. In conclusion, patient space turns out to be richer and more complex than current models suggest. TDA could become a powerful tool in a researcher's arsenal for interpreting high-dimensional data by providing novel insights into biological processes and improving our understanding of clinical and biological data sets.

A robust clustering strategy for stratification unveils unique patient subgroups in acutely decompensated cirrhosis

BACKGROUND

Patient heterogeneity poses significant challenges for managing individuals and designing clinical trials, especially in complex diseases. Existing classifications rely on outcome-predicting scores, potentially overlooking crucial elements contributing to heterogeneity without necessarily impacting prognosis.

METHODS

To address patient heterogeneity, we developed ClustALL, a computational pipeline that simultaneously faces diverse clinical data challenges like mixed types, missing values, and collinearity. ClustALL enables the unsupervised identification of patient stratifications while filtering for stratifications that are robust against minor variations in the population (population-based) and against limited adjustments in the algorithm's parameters (parameter-based).

RESULTS

Applied to a European cohort of patients with acutely decompensated cirrhosis (n = 766), ClustALL identified five robust stratifications, using only data at hospital admission. All stratifications included markers of impaired liver function and number of organ dysfunction or failure, and most included precipitating events. When focusing on one of these stratifications, patients were categorized into three clusters characterized by typical clinical features; notably, the 3-cluster stratification showed a prognostic value. Re-assessment of patient stratification during follow-up delineated patients' outcomes, with further improvement of the prognostic value of the stratification. We validated these findings in an independent prospective multicentre cohort of patients from Latin America (n = 580).

CONCLUSIONS

By applying ClustALL to patients with acutely decompensated cirrhosis, we identified three patient clusters. Following these clusters over time offers insights that could guide future clinical trial design. ClustALL is a novel and robust stratification method capable of addressing the multiple challenges of patient stratification in most complex diseases.

Supervised convex clustering

Clustering has long been a popular unsupervised learning approach to identify groups of similar objects and discover patterns from unlabeled data in many applications. Yet, coming up with meaningful interpretations of the estimated clusters has often been challenging precisely due to their unsupervised nature. Meanwhile, in many real-world scenarios, there are some noisy supervising auxiliary variables, for instance, subjective diagnostic opinions, that are related to the observed heterogeneity of the unlabeled data. By leveraging information from both supervising auxiliary variables and unlabeled data, we seek to uncover more scientifically interpretable group structures that may be hidden by completely unsupervised analyses. In this work, we propose and develop a new statistical pattern discovery method named supervised convex clustering (SCC) that borrows strength from both information sources and guides towards finding more interpretable patterns via a joint convex fusion penalty. We develop several extensions of SCC to integrate different types of supervising auxiliary variables, to adjust for additional covariates, and to find biclusters. We demonstrate the practical advantages of SCC through simulations and a case study on Alzheimer's disease genomics. Specifically, we discover new candidate genes as well as new subtypes of Alzheimer's disease that can potentially lead to better understanding of the underlying genetic mechanisms responsible for the observed heterogeneity of cognitive decline in older adults.

Transgenerational impact of climatic changes on cotton production

Changing climatic conditions are an increasing threat to cotton production worldwide. There is a need to develop multiple stress-tolerant cotton germplasms that can adapt to a wide range of environments. For this purpose, 30 cotton genotypes were evaluated for two years under drought (D), heat (H), and drought + heat stresses (DH) under field conditions. Results indicated that plant height, number of bolls, boll weight, seed cotton yield, fiber fineness, fiber strength, fiber length, K⁺, K⁺/Na⁺, relative water contents (RWC), chlorophyll a and b, carotenoids, and total soluble proteins got reduced under D and H and were lowest under DH, whereas superoxidase dismutase (SOD), H₂O₂, Na⁺, GOT%, total phenolic contents, ascorbate, and flavonoids got increased for consecutive years. Correlation studies indicated that there was a positive correlation between most of the traits, but a negative correlation with H₂O₂ and Na⁺ ions. PCA and clustering analysis indicated that MNH-786, KAHKSHAN, CEMB-33, MS-71, FH-142, NIAB-820, CRS-2007, and FH-312 consistently performed better than other genotypes for most traits under stress conditions. Identified genotypes can be utilized in the future cotton breeding program to develop high-yielding, climate change-resilient cotton.

Biochemical, physiological, and molecular characterisation of a large collection of aerobic endospore-forming bacteria isolated from Brazilian soils

The aerobic endospore-forming bacteria (AEFB) comprise species of Bacillus and related genera and have long been regarded as prominent constituents of the soil bacterial community. The wide diversity of AEFB renders appropriate categorisation and generalisations a challenging task. We previously isolated 312 AEFB strains from Brazilian soils that we designated SDF (Solo do Distrito Federal) strains. To better understand the SDF diversity and explore their biotechnological potential, we addressed the biochemical and physiological profiles of these 312 environmental strains by performing 30 tests in this work. Of these, the 16S rRNA gene sequences segregated 238 SDF strains into four genera in the family Bacillaceae and two in the Paenibacillaceae. Bacillus spp. were the most prevalent, followed by species of Paenibacillus. We summarised the phenotypic test relationships among selected SDF strains using a Pearson correlation-based clustering represented in heatmaps. In practice, biochemical and physiological profiles are often less discriminatory than molecular data and may be unstable because of the loss of traits. Although these test reactions are not universally positive or negative within species, they may define biotypes and be efficient strain markers, enhancing the accuracy of unknown sample identification. It can also help select the most representative phenotypes of samples. Along with the other phenotypic and genotypic data, the present results are of great importance for the robust classification of the SDF strains within the scope of the polyphasic approach.

MIGR: A Categorical Data Clustering Algorithm Based on Information Gain in Rough Set Theory

Clustering techniques are used to split data into clusters where each cluster contains elements that look more similar to elements in the same cluster than elements in other clusters. Some of these techniques are capable of handling clustering process uncertainty, while other techniques may have stability issues. In this paper, a novel method, called Minimum Information Gain Roughness (MIGR), is proposed to select the clustering attribute based on information entropy with rough set theory. To evaluate its performance, three benchmark UCI datasets are chosen to be clustered by using MIGR. Then, the resulting clusters are compared to those which are resulted from applying Min-Min-Rough (MMR) and information-theoretic dependency roughness (ITDR) algorithms. Both last-mentioned techniques were already compared with a variety of clustering algorithms like k-modes, fuzzy centroids, and fuzzy k-modes. The Global purity, the overall purity, and F-measure are considered here as performance measures to compare the quality of the resulting clusters. The experimental results show that the MIGR algorithm outperforms both MMR and ITDR algorithms for clustering categorical data.

Heterogeneity of patients with functional/dissociative seizures: Three multidimensional profiles

OBJECTIVE

Current concepts highlight the neurological and psychological heterogeneity of functional/dissociative seizures (FDS). However, it remains uncertain whether it is possible to distinguish between a limited number of subtypes of FDS disorders. We aimed to identify profiles of distinct FDS subtypes by cluster analysis of a multidimensional dataset without any a priori hypothesis.

METHODS

We conducted an exploratory, prospective multicenter study of 169 patients with FDS. We collected biographical, trauma (childhood and adulthood traumatic experiences), semiological (seizure characteristics), and psychopathological data (psychiatric comorbidities, dissociation, and alexithymia) through psychiatric interviews and standardized scales. Clusters were identified by the Partitioning Around Medoids method. The similarity of patients was computed using Gower distance. The clusters were compared using analysis of variance, chi-squared, or Fisher exact tests.

RESULTS

Three patient clusters were identified in this exploratory, hypothesis-generating study and named on the basis of their most prominent characteristics: A "No/Single Trauma" group (31.4%), with more male patients, intellectual disabilities, and nonhyperkinetic seizures, and a low level of psychopathology; A "Cumulative Lifetime Traumas" group (42.6%), with clear female predominance, hyperkinetic seizures, relatively common comorbid epilepsy, and a high level of psychopathology; and A "Childhood Traumas" group (26%), commonly with comorbid epilepsy, history of childhood sexual abuse (75%), and posttraumatic stress disorder, but also with a high level of anxiety and dissociation.

SIGNIFICANCE

Although our cluster analysis was undertaken without any a priori hypothesis, the nature of the trauma history emerged as the most important differentiator between three common FDS disorder subtypes. This subdifferentiation of FDS disorders may facilitate the development of more specific therapeutic programs for each patient profile.

UL34 Deletion Restricts Human Cytomegalovirus Capsid Formation and Maturation

Over 50% of the world’s population is infected with Human Cytomegalovirus (HCMV). HCMV is responsible for serious complications in the immuno-compromised and is a leading cause of congenital birth defects. The molecular function of many HCMV proteins remains unknown, and a deeper understanding of the viral effectors that modulate virion maturation is required. In this study, we observed that UL34 is a viral protein expressed with leaky late kinetics that localises to the nucleus during infection. Deletion of UL34 from the HCMV genome (ΔUL34) did not abolish the spread of HCMV. Instead, over >100-fold fewer infectious virions were produced, so we report that UL34 is an augmenting gene. We found that ΔUL34 is dispensable for viral DNA replication, and its absence did not alter the expression of IE1, MCP, gB, UL26, UL83, or UL99 proteins. In addition, ΔUL34 infections were able to progress through the replication cycle to form a viral assembly compartment; however, virion maturation in the cytoplasm was abrogated. Further examination of the nucleus in ΔUL34 infections revealed replication compartments with aberrant morphology, containing significantly less assembled capsids, with almost none undergoing subsequent maturation. Therefore, this work lays the foundation for UL34 to be further investigated in the context of nuclear organization and capsid maturation during HCMV infection.

A retrotransposon storm marks clinical phenoconversion to late-onset Alzheimer's disease

Recent reports have suggested that the reactivation of otherwise transcriptionally silent transposable elements (TEs) might induce brain degeneration, either by dysregulating the expression of genes and pathways implicated in cognitive decline and dementia or through the induction of immune-mediated neuroinflammation resulting in the elimination of neural and glial cells. In the work we present here, we test the hypothesis that differentially expressed TEs in blood could be used as biomarkers of cognitive decline and development of AD. To this aim, we used a sample of aging subjects (age > 70) that developed late-onset Alzheimer’s disease (LOAD) over a relatively short period of time (12–48 months), for which blood was available before and after their phenoconversion, and a group of cognitive stable subjects as controls. We applied our developed and validated customized pipeline that allows the identification, characterization, and quantification of the differentially expressed (DE) TEs before and after the onset of manifest LOAD, through analyses of RNA-Seq data. We compared the level of DE TEs within more than 600,000 TE-mapping RNA transcripts from 25 individuals, whose specimens we obtained before and after their phenotypic conversion (phenoconversion) to LOAD, and discovered that 1790 TE transcripts showed significant expression differences between these two timepoints (logFC ± 1.5, logCMP > 5.3, nominal p value < 0.01). These DE transcripts mapped both over- and under-expressed TE elements. Occurring before the clinical phenoconversion, this TE storm features significant increases in DE transcripts of LINEs, LTRs, and SVAs, while those for SINEs are significantly depleted. These dysregulations end with signs of manifest LOAD. This set of highly DE transcripts generates a TE transcriptional profile that accurately discriminates the before and after phenoconversion states of these subjects. Our findings suggest that a storm of DE TEs occurs before phenoconversion from normal cognition to manifest LOAD in risk individuals compared to controls, and may provide useful blood-based biomarkers for heralding such a clinical transition, also suggesting that TEs can indeed participate in the complex process of neurodegeneration.

Clinical features and treatment response to differentiate idiopathic peritonitis from non-strangulating intestinal infarction of the pelvic flexure associated with Strongylus vulgaris infection in the horse

Background

Peritonitis in horses secondary to non-strangulating infarction (NSII) has a guarded prognosis, even after intestinal resection. In contrast, horses with idiopathic peritonitis respond well to medical treatment. Affected horses in both cases often show signs of both colic and systemic inflammation, but early diagnosis is crucial for optimal treatment and an accurate prognosis. One cause of NSII is thrombus formation secondary to Strongylus vulgaris larval migration. There has been a documented increase in S. vulgaris prevalence in Sweden since the implementation of selective anthelmintic treatment in 2007, which subsequently could result in a rise in NSII cases. In a retrospective clinical study, medical records from cases diagnosed with NSII of the pelvic flexure or idiopathic peritonitis from three equine referral hospitals in Sweden during 2017–2020 were reviewed. Information including demographic data, relevant medical history, and clinical- and laboratory parameters were obtained from patient records. To facilitate the differentiation between cases of idiopathic peritonitis and cases with confirmed NSII of the pelvic flexure, the aim of the study was to compare clinical and laboratory parameters, clinical progression and initial response to antimicrobial treatment. A secondary aim was to compare survival-rates.

Results

Horses with NSII (n = 20) were significantly more likely to present during the winter months with a poorer response to medical treatment within 48 h. Cases of idiopathic peritonitis (n = 107) had a 100% survival rate with medical treatment, although one case required surgical correction of a colon displacement. In comparison, all confirmed NSII cases were non-responsive to antimicrobial treatment, with a survival rate to discharge of 50% after colon resection. Specific rectal findings and peripheral blood neutropenia were strongly associated with NSII.

Conclusions

In Sweden, idiopathic peritonitis cases still predominate over S. vulgaris associated NSII cases and have an excellent survival rate with antimicrobial treatment. However, horses presenting with septic peritonitis during the winter months with a palpable rectal mass and displaying fever and colic signs beyond 48 h of medical treatment are likely to suffer from NSII of the pelvic flexure and should be considered for abdominal surgery.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12917-022-03248-x.

The Vulvar Immunohistochemical Panel (VIP) Project: Molecular Profiles of Vulvar Squamous Cell Carcinoma

Simple Summary

This study investigated the immunohistochemical expression of 14 biological markers as potential prognostic/therapeutic factors in vulvar squamous cell carcinoma, comparing 53 node-negative (Group A) and 48 node-positive (Group B) patients. Our results show a significantly higher p16 expression (surrogate of HPV-related tumors) in the vulvar samples of non-metastatic patients. In Group B, PD-L1 positivity and high EGFR expression were found in the vast majority of vulvar and/or nodal specimens. VEGF showed strong/moderate-diffuse expression in almost 14% of all vulvar samples. A mutated p53 and over-expressed PD-L1 showed a significant association with nodal metastasis. Our results support a potential role of immune checkpoint inhibitors and anti-VEGF and anti-EGFR drugs, especially in patients with worse prognosis (metastatic, HPV-independent). A panel including EGFR, VEGF, PDL1, p16, and p53 might be performed routinely in primary tumor and repeated in case of lymph node metastases to identify changes in marker expression.

Abstract

Introduction: The study’s aim was to investigate the immunohistochemical (IHC) expression of biological markers as potential prognostic/therapeutic factors in vulvar squamous cell carcinoma (VSCC). Methodology: A series of 101 patients surgically treated at our center from 2016 to 2020 were retrospectively enrolled: 53 node-negative (Group A) and 48 node-positive (Group B). A total of 146 samples, 101 from primary tumor (T) and 45 from nodal metastases (N), were investigated. The IHC panel included: p16, p53, MLH1, MSH2, MSH6, PMS2, PD-L1, CD3, HER2/neu, ER, PR, EGFR, VEGF, and CD31. The reactions were evaluated on qualitative and semi-quantitative scales. Generalized Linear Model (GLM) and cluster analysis were performed in R statistical environment. A distance plot compared the IHC panel of T with the correspondent N. Results: In Group A: p16-positive expression (surrogate of HPV-dependent pathway) was significantly higher (20.8% vs. 6.2%, p = 0.04). In Group B: PD-L1 positivity and high EGFR expression were found, respectively, in 77.1% and 97.9% patients (T and/or N). Overall, p16-negative tumors showed a higher PD-L1 expression (60.9% vs. 50.0%). In both groups: tumoral immune infiltration (CD3 expression) was mainly moderate/intense (80% vs. 95%); VEGF showed strong/moderate-diffuse expression in 13.9% of T samples; CD31, related to tumoral microvessel density (MVD), showed no difference between groups; a mutated p53 and over-expressed PD-L1 showed significant association with nodal metastasis, with Odds Ratios (OR) of 4.26 (CI 95% = 1.14–15.87, p = 0.03) and 2.68 (CI 95% = 1.0–7.19, p < 0.05), respectively; since all mismatch repair proteins (MMR) showed a retained expression and ER, PR, and HER2/neu were negative, they were excluded from further analysis. The cluster analysis identified three and four sub-groups of molecular profiles, respectively, in Group A and B, with no difference in prognosis. The molecular signature of each N and corresponding T diverged significantly in 18/41 (43.9%) cases. Conclusions: Our results support a potential role of immune checkpoint inhibitors and anti-VEGF and anti-EGFR drugs especially in patients with worse prognosis (metastatic, HPV-independent). A panel including EGFR, VEGF, PDL1, p16, and p53 might be performed routinely in primary tumor and repeated in case of lymph node metastases to identify changes in marker expression.

Unraveling the Complexity to Observe Associations Between Welfare Indicators and Hair Cortisol Concentration in Dairy Calves

Using levels of the stress hormone cortisol as an indicator for welfare is a common, but debated practice. In this observational study, hair cortisol concentration (HCC) of samples from 196 dairy calves from 7 to 302 days of age collected from 12 Swedish farms was determined using a commercially available ELISA. An assessment of animal welfare, assessed using animal-based indicators, was performed on the day of sampling. First, methodological factors with the potential to impact HCC and the effect of age were analyzed using generalized additive models. This revealed a significant peak in hair cortisol in young calves (around 50 days of age) and an association between fecal contamination of hair samples and the level of cortisol extracted. Second, associations between welfare indicators and HCC were explored using cluster analysis and regularized regression. The results show a complex pattern, possibly related to different coping styles of the calves, and indicators of poor welfare were associated with both increased and decreased hair cortisol levels. High cortisol levels were associated with potential indicators of competition, while low cortisol levels were associated with the signs of poor health or a poor environment. When running the regularized regression analysis without the contaminated hair samples and with the contaminated samples (including a contamination score), the results did not change, indicating that it may be possible to use a contamination score to correct for contamination.

Antimicrobial resistance in wildlife and in the built environment in a wildlife rehabilitation center

Injured and orphaned wildlife are often brought to Wildlife Rehabilitation Centers (WRC) to be cared for by professionals to ultimately be released back to their natural habitats. In these centers, animals may spend months and frequently receive prolonged antibiotic therapy. Therefore, WRC may play a role in the emergence and dissemination of antimicrobial resistance (AMR). The goal of this study was to investigate the presence and antibiotic resistance profiles of Gram-negative bacteria with reduced susceptibility to cephalosporins in both the wildlife admitted to a WRC and in the WRC built environment in Chile. A cross-sectional study was conducted sampling animals undergoing rehabilitation (n = 64) and the WRC environment (n = 160). Isolated bacterial species were identified with MALDI-TOF, and antimicrobial susceptibility determined using the disk diffusion method. Enterobacteriaceae and Pseudomonadaceae were the dominant bacterial families among the environmental (n = 78) and animal (n = 31) isolates. For Enterobacteriaceae, isolates of the most abundant species (E. coli) were classified into 20 antibiotic resistance profiles, with eight of those isolates being resistant to more than nine antibiotics, including imipenem. Isolates of the Pseudomonadaceae family identified 11 isolates with resistance to antibiotics such as carbapenems and quinolones. Even though a cluster analysis based on antibiotic resistance patterns did not show a clear overlap between environmental and animal isolates, it is important to highlight the identification of isolates resistant to carbapenems, which is very relevant from a public health perspective. Further, numerous antibiotic resistance profiles were observed in different bacterial species, indicating not only environmental contamination with a wide diversity of bacteria, but also a wide diversity of resistant bacteria in animals at the WRC. The approach taken by sampling animals and their hospital environment can be useful in understanding AMR dynamics in wildlife rehabilitation settings, as well as the potential dissemination of AMR into the natural environment.

•

The wildlife center was contaminated with wide diversity of resistant bacteria.

•

There was wide diversity of resistant bacteria in wildlife at the center.

•

Resistant isolates to carbapenems were present, which has public health relevance.

•

No clear overlap between wildlife and the center antibiotic resistance patterns.

•

Wildlife rehabilitation should be considered in antimicrobial resistance dynamics.

A phenomapping-derived tool to personalize the selection of anatomical vs. functional testing in evaluating chest pain (ASSIST)

AIMS

Coronary artery disease is frequently diagnosed following evaluation of stable chest pain with anatomical or functional testing. A more granular understanding of patient phenotypes that benefit from either strategy may enable personalized testing.

METHODS AND RESULTS

Using participant-level data from 9572 patients undergoing anatomical (n = 4734) vs. functional (n = 4838) testing in the PROMISE (PROspective Multicenter Imaging Study for Evaluation of Chest Pain) trial, we created a topological representation of the study population based on 57 pre-randomization variables. Within each patient's 5% topological neighbourhood, Cox regression models provided individual patient-centred hazard ratios for major adverse cardiovascular events and revealed marked heterogeneity across the phenomap [median 1.11 (10th to 90th percentile: 0.52-2.61]), suggestive of distinct phenotypic neighbourhoods favouring anatomical or functional testing. Based on this risk phenomap, we employed an extreme gradient boosting algorithm in 80% of the PROMISE population to predict the personalized benefit of anatomical vs. functional testing using 12 model-derived, routinely collected variables and created a decision support tool named ASSIST (Anatomical vs. Stress teSting decIsion Support Tool). In both the remaining 20% of PROMISE and an external validation set consisting of patients from SCOT-HEART (Scottish COmputed Tomography of the HEART Trial) undergoing anatomical-first vs. functional-first assessment, the testing strategy recommended by ASSIST was associated with a significantly lower incidence of each study's primary endpoint (P = 0.0024 and P = 0.0321 for interaction, respectively), as well as a harmonized endpoint of all-cause mortality or non-fatal myocardial infarction (P = 0.0309 and P < 0.0001 for interaction, respectively).

CONCLUSION

We propose a novel phenomapping-derived decision support tool to standardize the selection of anatomical vs. functional testing in the evaluation of stable chest pain, validated in two large and geographically diverse clinical trial populations.

Conceptualizing problems with symptoms, function, health behavior, health-seeking skills, and financial strain in breast cancer survivors using hierarchical clustering

Purpose

Determine whether a diverse set of problems experienced by breast cancer survivors (BCS) following curative treatment can be formulated into a reduced number of clusters, potentially simplifying the conceptualization of these problems.

Method

Female BCS were recruited from four cancer hospitals in China. The Chinese translation of the Cancer Survivor Profile (CSPro) was used to measure 18 common problem areas, as supported by epidemiological and phenomenological research. The Functional Assessment of Cancer Therapy–Breast (FACT-B) was used to measure quality of life, as a validation of any observed groupings. Hierarchical clustering using multiple distance criteria and aggregation methods to detect patterns of problems was used.

Results

A total of 1008 BCS (mean 46.51 years old) living in both urban and rural areas were investigated. Hierarchical cluster analysis identified two major clusters of problems. One set was classified as “functional limitations,” while the other cluster was labeled “multi-problems.” Those who fell into the multi-problem cluster experienced poorer quality of life.

Conclusion

Eighteen non-medical problems were broken down into two major clusters: (1) limitations in higher level functions required of daily life and (2) limitations in health care–seeking skills, problems with certain symptoms, unhealthy behaviors, and financial problems related to cancer. The breakdown of problem areas into these two clusters may help identify common mechanisms.

Implications for Cancer Survivors

In the future, the search for common clusters and the mechanisms for the many problems that breast cancer survivors and other cancer survivors can experience following primary treatment may improve how we help manage these problems in the future.

Transforming a Patient Registry Into a Customized Data Set for the Advanced Statistical Analysis of Health Risk Factors and for Medication-Related Hospitalization Research: Retrospective Hospital Patient Registry Study

BACKGROUND

Hospital patient registries provide substantial longitudinal data sets describing the clinical and medical health statuses of inpatients and their pharmacological prescriptions. Despite the multiple advantages of routinely collecting multidimensional longitudinal data, those data sets are rarely suitable for advanced statistical analysis and they require customization and synthesis.

OBJECTIVE

The aim of this study was to describe the methods used to transform and synthesize a raw, multidimensional, hospital patient registry data set into an exploitable database for the further investigation of risk profiles and predictive and survival health outcomes among polymorbid, polymedicated, older inpatients in relation to their medicine prescriptions at hospital discharge.

METHODS

A raw, multidimensional data set from a public hospital was extracted from the hospital registry in a CSV (.csv) file and imported into the R statistical package for cleaning, customization, and synthesis. Patients fulfilling the criteria for inclusion were home-dwelling, polymedicated, older adults with multiple chronic conditions aged ≥65 who became hospitalized. The patient data set covered 140 variables from 20,422 hospitalizations of polymedicated, home-dwelling older adults from 2015 to 2018. Each variable, according to type, was explored and computed to describe distributions, missing values, and associations. Different clustering methods, expert opinion, recoding, and missing-value techniques were used to customize and synthesize these multidimensional data sets.

RESULTS

Sociodemographic data showed no missing values. Average age, hospital length of stay, and frequency of hospitalization were computed. Discharge details were recoded and summarized. Clinical data were cleaned up and best practices for managing missing values were applied. Seven clusters of medical diagnoses, surgical interventions, somatic, cognitive, and medicines data were extracted using empirical and statistical best practices, with each presenting the health status of the patients included in it as accurately as possible. Medical, comorbidity, and drug data were recoded and summarized.

CONCLUSIONS

A cleaner, better-structured data set was obtained, combining empirical and best-practice statistical approaches. The overall strategy delivered an exploitable, population-based database suitable for an advanced analysis of the descriptive, predictive, and survival statistics relating to polymedicated, home-dwelling older adults admitted as inpatients. More research is needed to develop best practices for customizing and synthesizing large, multidimensional, population-based registries.

INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID)

RR2-10.1136/bmjopen-2019-030030.

Exploration of Multiparameter Hematoma 3D Image Analysis for Predicting Outcome After Intracerebral Hemorrhage

BACKGROUND

Rapid diagnosis and proper management of intracerebral hemorrhage (ICH) play a crucial role in the outcome. Prediction of the outcome with a high degree of accuracy based on admission data including imaging information can potentially influence clinical decision-making practice.

METHODS

We conducted a retrospective multicenter study of consecutive ICH patients admitted between 2012-2017. Medical history, admission data, and initial head computed tomography (CT) scan were collected. CT scans were semiautomatically segmented for hematoma volume, hematoma density histograms, and sphericity index (SI). Discharge unfavorable outcomes were defined as death or severe disability (modified Rankin Scores 4-6). We compared (1) hematoma volume alone; (2) multiparameter imaging data including hematoma volume, location, density heterogeneity, SI, and midline shift; and (3) multiparameter imaging data with clinical information available on admission for ICH outcome prediction. Multivariate analysis and predictive modeling were used to determine the significance of hematoma characteristics on the outcome.

RESULTS

We included 430 subjects in this analysis. Models using automated hematoma segmentation showed incremental predictive accuracies for in-hospital mortality using hematoma volume only: area under the curve (AUC): 0.85 [0.76-0.93], multiparameter imaging data (hematoma volume, location, CT density, SI, and midline shift): AUC: 0.91 [0.86-0.97], and multiparameter imaging data plus clinical information on admission (Glasgow Coma Scale (GCS) score and age): AUC: 0.94 [0.89-0.99]. Similarly, severe disability predictive accuracy varied from AUC: 0.84 [0.76-0.93] for volume-only model to AUC: 0.88 [0.80-0.95] for imaging data models and AUC: 0.92 [0.86-0.98] for imaging plus clinical predictors.

CONCLUSIONS

Multiparameter models combining imaging and admission clinical data show high accuracy for predicting discharge unfavorable outcome after ICH.

Integrative Generalized Convex Clustering Optimization and Feature Selection for Mixed Multi-View Data

In mixed multi-view data, multiple sets of diverse features are measured on the same set of samples. By integrating all available data sources, we seek to discover common group structure among the samples that may be hidden in individualistic cluster analyses of a single data view. While several techniques for such integrative clustering have been explored, we propose and develop a convex formalization that enjoys strong empirical performance and inherits the mathematical properties of increasingly popular convex clustering methods. Specifically, our Integrative Generalized Convex Clustering Optimization (iGecco) method employs different convex distances, losses, or divergences for each of the different data views with a joint convex fusion penalty that leads to common groups. Additionally, integrating mixed multi-view data is often challenging when each data source is high-dimensional. To perform feature selection in such scenarios, we develop an adaptive shifted group-lasso penalty that selects features by shrinking them towards their loss-specific centers. Our so-called iGecco+ approach selects features from each data view that are best for determining the groups, often leading to improved integrative clustering. To solve our problem, we develop a new type of generalized multi-block ADMM algorithm using sub-problem approximations that more efficiently fits our model for big data sets. Through a series of numerical experiments and real data examples on text mining and genomics, we show that iGecco+ achieves superior empirical performance for high-dimensional mixed multi-view data.

Assessing invasiveness of subsolid lung adenocarcinomas with combined attenuation and geometric feature models

The aim of this study was to develop and test multiclass predictive models for assessing the invasiveness of individual lung adenocarcinomas presenting as subsolid nodules on computed tomography (CT). 227 lung adenocarcinomas were included: 31 atypical adenomatous hyperplasia and adenocarcinomas in situ (class H1), 64 minimally invasive adenocarcinomas (class H2) and 132 invasive adenocarcinomas (class H3). Nodules were segmented, and geometric and CT attenuation features including functional principal component analysis features (FPC1 and FPC2) were extracted. After a feature selection step, two predictive models were built with ordinal regression: Model 1 based on volume (log) (logarithm of the nodule volume) and FPC1, and Model 2 based on volume (log) and Q.875 (CT attenuation value at the 87.5% percentile). Using the 200-repeats Monte-Carlo cross-validation method, these models provided a multiclass classification of invasiveness with discriminative power AUCs of 0.83 to 0.87 and predicted the class probabilities with less than a 10% average error. The predictive modelling approach adopted in this paper provides a detailed insight on how the value of the main predictors contribute to the probability of nodule invasiveness and underlines the role of nodule CT attenuation features in the nodule invasiveness classification.

Harnessing Population Pedigree Data and Machine Learning Methods to Identify Patterns of Familial Bladder Cancer Risk

BACKGROUND

Relatives of patients with bladder cancer have been shown to be at increased risk for kidney, lung, thyroid, and cervical cancer after correcting for smoking-related behaviors that may concentrate in some families. We demonstrate a novel approach to simultaneously assess risks for multiple cancers to identify distinct multicancer configurations (multiple different cancer types that cluster in relatives) surrounding patients with familial bladder cancer.

METHODS

This study takes advantage of a unique population-level data resource, the Utah Population Database (UPDB), containing vast genealogy and statewide cancer data. Familial risk is measured using standardized incidence risk (SIR) ratios that account for sex, age, birth cohort, and person-years of the pedigree members.

RESULTS

We identify 1,023 families with a significantly higher bladder cancer rate than population controls (familial bladder cancer). Familial SIRs are then calculated across 25 cancer types, and a weighted Gower distance with K-medoids clustering is used to identify familial multicancer configurations (FMC). We found five FMCs, each exhibiting a different pattern of cancer aggregation. Of the 25 cancer types studied, kidney and prostate cancers were most commonly enriched in the familial bladder cancer clusters. Laryngeal, lung, stomach, acute lymphocytic leukemia, Hodgkin disease, soft-tissue carcinoma, esophageal, breast, lung, uterine, thyroid, and melanoma cancers were the other cancer types with increased incidence in familial bladder cancer families.

CONCLUSIONS

This study identified five familial bladder cancer FMCs showing unique risk patterns for cancers of other organs, suggesting phenotypic heterogeneity familial bladder cancer.

IMPACT

FMC configurations could permit better definitions of cancer phenotypes (subtypes or multicancer) for gene discovery and environmental risk factor studies.

Socially engaged calves are more likely to be colonised by VTEC O157:H7 than individuals showing signs of poor welfare

In cattle herds, the transmission and persistence of VTEC O157:H7 (a serotype of verotoxin-producing Escherichia coli – known for its life threatening complications in humans) is dependent on a small proportion of cattle who become colonised and shed high numbers of the bacteria. Reducing the proportion of these animals is considered key for decreasing the prevalence of VTEC O157:H7. In this study, observations of calf behaviour and animal-based welfare indicators were used to explore individual risk factors and underlying drivers of colonisation in Swedish dairy calves. Interdependencies between variables led to three different approaches being used to visualize and explore the associations. Combining the results of all methods revealed similar patterns and suggest that healthy animals, actively grooming and interacting with others calves in the group have a higher risk of colonisation than small dairy calves in poor condition (diarrhoea, poor ruminal fill, poor body condition score and nasal discharge). This lends no support to the hypothesis that reduced welfare is a risk factor for VTEC O157:H7, but implies that individual differences in calf behaviour affect oral exposure to the bacteria so driving the risk of colonisation. This new finding has important implications for understanding of VTEC O157:H7 transmission within farms.

Cytogenetic intraclonal heterogeneity of plasma cell dyscrasia in AL amyloidosis as compared with multiple myeloma

Analysis of intraclonal heterogeneity has yielded insights into the clonal evolution of hematologic malignancies. We compared the clonal and subclonal compositions of the underlying plasma cell dyscrasia in 544 systemic light chain amyloidosis (PC-AL) patients with 519 patients with monoclonal gammopathy of undetermined significance (MGUS), smoldering multiple myeloma (SMM), or symptomatic MM; ie, PC-non-AL patients). Using interphase fluorescence in situ hybridization, subclones were stringently defined as clone size below two thirds of the largest clone and an absolute difference of ≥30%. Subclones were found less frequently in the PC-AL group, at 199 (36.6%) of 544 as compared with 267 (51.4%) of 519 in the PC-non-AL group (P < .001), and were not associated with the stage of plasma cell dyscrasia in either entity. In both groups, translocation t(11;14), other immunoglobulin heavy chain translocations, and hyperdiploidy were typically found as main clones, whereas gain of 1q21 and deletions of 8p21, 13q14, and 17p13 were frequently found as subclones. There were no shifts in the subclone/main clone ratio depending on the MGUS, SMM, or MM stage of plasma cell dyscrasia. In multivariate analysis, t(11;14) was associated with lower rates of subclone formation and hyperdiploidy with higher rates. PC-AL itself lost statistical significance, demonstrating that the lower subclone frequency in AL is a reflection of its exceptionally high t(11;14) frequency. In summary, the subclone patterns in PC-AL and PC-non-AL are closely related, implying that subclone formation depends on the main cytogenetic categories and is independent of disease entity and stage.

Identification of an early diagnostic biomarker of lung adenocarcinoma based on co-expression similarity and construction of a diagnostic model

Background

The purpose of this study was to achieve early and accurate diagnosis of lung cancer and long-term monitoring of the therapeutic response.

Methods

We downloaded GSE20189 from GEO database as analysis data. We also downloaded human lung adenocarcinoma RNA-seq transcriptome expression data from the TCGA database as validation data. Finally, the expression of all of the genes underwent z test normalization. We used ANOVA to identify differentially expressed genes specific to each stage, as well as the intersection between them. Two methods, correlation analysis and co-expression network analysis, were used to compare the expression patterns and topological properties of each stage. Using the functional quantification algorithm, we evaluated the functional level of each significantly enriched biological function under different stages. A machine-learning algorithm was used to screen out significant functions as features and to establish an early diagnosis model. Finally, survival analysis was used to verify the correlation between the outcome and the biomarkers that we found.

Results

We screened 12 significant biomarkers that could distinguish lung cancer patients with diverse risks. Patients carrying variations in these 12 genes also presented a poor outcome in terms of survival status compared with patients without variations.

Conclusions

We propose a new molecular-based noninvasive detection method. According to the expression of the stage-specific gene set in the peripheral blood of patients with lung cancer, the difference in the functional level is quantified to realize the early diagnosis and prediction of lung cancer.