Why most gene expression signatures of tumors have not been useful in the clinic

Identifying high-risk multiple myeloma patients: A novel approach using a clonal gene signature

Multiple myeloma (MM) is a heterogeneous disease with a small subset of high-risk patients having poor prognoses. Identifying these patients is crucial for treatment management and strategic decisions. In this study, we developed a novel computational framework to define prognostic gene signatures by selecting genes with expression driven by clonal copy number alterations. We applied this framework to MM and developed a clonal gene signature (CGS) consisting of 22 genes and evaluated in five independent datasets. The CGS provided significant prognostic values after adjusting for well-established factors including cytogenetic abnormalities, International Staging System (ISS), and Revised ISS (R-ISS). Importantly, CGS demonstrated higher performance in identifying high-risk patients compared to the GEP70 and SKY92 signatures recommended for prognostic stratification of MM. CGS can further stratify patients into subgroups with significantly differential prognoses when applied to the high- and low-risk groups identified by GEP70 and SKY92. Additionally, CGS scores are significantly associated with patient response to dexamethasone, a commonly used treatment for MM. In summary, we proposed a computational framework that requires only gene expression data to identify CGSs for prognosis prediction. CGS provides a useful biomarker for improving prognostic stratification in MM, especially for identifying the highest-risk patients.

DNA-based molecular classifiers for the profiling of gene expression signatures

Although gene expression signatures offer tremendous potential in diseases diagnostic and prognostic, but massive gene expression signatures caused challenges for experimental detection and computational analysis in clinical setting. Here, we introduce a universal DNA-based molecular classifier for profiling gene expression signatures and generating immediate diagnostic outcomes. The molecular classifier begins with feature transformation, a modular and programmable strategy was used to capture relative relationships of low-concentration RNAs and convert them to general coding inputs. Then, competitive inhibition of the DNA catalytic reaction enables strict weight assignment for different inputs according to their importance, followed by summation, annihilation and reporting to accurately implement the mathematical model of the classifier. We validated the entire workflow by utilizing miRNA expression levels for the diagnosis of hepatocellular carcinoma (HCC) in clinical samples with an accuracy 85.7%. The results demonstrate the molecular classifier provides a universal solution to explore the correlation between gene expression patterns and disease diagnostics, monitoring, and prognosis, and supports personalized healthcare in primary care.

Identification of biomarkers for the early detection of non-small cell lung cancer: a systematic review and meta-analysis

Lung cancer (LC) causes few symptoms in the earliest stages, leading to one of the highest mortality rates among cancers. Low-dose computerised tomography (LDCT) is used to screen high-risk individuals, reducing the mortality rate by 20%. However, LDCT results in a high number of false positives and is associated with unnecessary follow-up and cost. Biomarkers with high sensitivities and specificities could assist in the early detection of LC, especially in patients with high-risk features. Carcinoembryonic antigen (CEA), cytokeratin 19 fragments and cancer antigen 125 have been found to be highly expressed during the later stages of LC but have low sensitivity in the earliest stages. We determined the best biomarkers for the early diagnosis of LC, using a systematic review of eight databases. We identified 98 articles that focussed on the identification and assessment of diagnostic biomarkers and achieved a pooled area under curve of 0.85 (95% CI 0.82-0.088), indicating that the diagnostic performance of these biomarkers when combined was excellent. Of the studies, 30 focussed on single/antigen panels, 22 on autoantibodies, 31 on miRNA and RNA panels, and 15 suggested the use of circulating DNA combined with CEA or neuron-specific enolase (NSE) for early LC detection. Verification of blood biomarkers with high sensitivities (Ciz1, exoGCC2, ITGA2B), high specificities (CYFR21-1, antiHE4, OPNV) or both (HSP90α, CEA) along with miR-15b and miR-27b/miR-21 from sputum may improve early LC detection. Further assessment is needed using appropriate sample sizes, control groups that include patients with non-malignant conditions, and standardised cut-off levels for each biomarker.

Advancing precision prognostication in neuro-oncology: Machine learning models for data-driven personalized survival predictions in IDH-wildtype glioblastoma

Background

Glioblastoma (GBM) remains associated with a dismal prognoses despite standard therapies. While population-level survival statistics are established, generating individualized prognosis remains challenging. We aim to develop machine learning (ML) models that generate personalized survival predictions for GBM patients to enhance prognostication.

Methods

Adult patients with histologically confirmed IDH-wildtype GBM from the National Cancer Database (NCDB) were analyzed. ML models were developed with TabPFN, TabNet, XGBoost, LightGBM, and Random Forest algorithms to predict mortality at 6, 12, 18, and 24 months postdiagnosis. SHapley Additive exPlanations (SHAP) were employed to enhance the interpretability of the models. Models were primarily evaluated using the area under the receiver operating characteristic (AUROC) values, and the top-performing models indicated by the highest AUROCs for each outcome were deployed in a web application that was created for individualized predictions.

Results

A total of 7537 patients were retrieved from the NCDB. Performance evaluation revealed the top-performing models for each outcome were built using the TabPFN algorithm. The TabPFN models yielded mean AUROCs of 0.836, 0.78, 0.732, and 0.724 in predicting 6, 12, 18, and 24 month mortality, respectively.

Conclusions

This study establishes ML models tailored to individual patients to enhance GBM prognostication. Future work should focus on external validation and dynamic updating as new data emerge.

Re-evaluation of publicly available gene-expression databases using machine-learning yields a maximum prognostic power in breast cancer

Gene expression signatures refer to patterns of gene activities and are used to classify different types of cancer, determine prognosis, and guide treatment decisions. Advancements in high-throughput technology and machine learning have led to improvements to predict a patient's prognosis for different cancer phenotypes. However, computational methods for analyzing signatures have not been used to evaluate their prognostic power. Contention remains on the utility of gene expression signatures for prognosis. The prevalent approaches include random signatures, expert knowledge, and machine learning to construct an improved signature. We unify these approaches to evaluate their prognostic power. Re-evaluation of publicly available gene-expression data from 8 databases with 9 machine-learning models revealed previously unreported results. Gene-expression signatures are confirmed to be useful in predicting a patient's prognosis. Convergent evidence from [Formula: see text] 10,000 signatures implicates a maximum prognostic power. By calculating the concordance index, which measures how well patients with different prognoses can be discriminated, we show that a signature can correctly discriminate patients' prognoses no more than 80% of the time. Additionally, we show that more than 50% of the potentially available information is still missing at this value. We surmise that an accurate prognosis must incorporate molecular, clinical, histological, and other complementary factors.

Transcriptomics in Tumor and Normal Lung Tissues Identify Patients With Early-Stage Non-Small-Cell Lung Cancer With High Risk of Postsurgery Recurrence Who May Benefit From Adjuvant Therapies

PURPOSE

The prognosis of patients with non-small-cell lung cancer (NSCLC), traditionally determined by anatomic histology and TNM staging, neglects the biological features of the tumor that may be important in determining patient outcome and guiding therapeutic interventions. Identifying patients with NSCLC at increased risk of recurrence after curative-intent surgery remains an important unmet need so that known effective adjuvant treatments can be offered to those at highest risk of recurrence.

METHODS

Relative gene expression level in the primary tumor and normal bronchial tissues was used to retrospectively assess their association with disease-free survival (DFS) in a cohort of 120 patients with NSCLC who underwent curative-intent surgery.

RESULTS

Low versus high Digital Display Precision Predictor (DDPP) score (a measure of relative gene expression) was significantly associated with shorter DFS (highest recurrence risk; P = .006) in all patients and in patients with TNM stages 1-2 (P = .00051; n = 83). For patients with stages 1-2 and low DDPP score (n = 29), adjuvant chemotherapy was associated with improved DFS (P = .0041). High co-overexpression of CTLA-4, PD-L1, and ICOS in normal lung (28 of 120 patients) was also significantly associated with decreased DFS (P = .0013), suggesting an immune tolerance to tumor neoantigens in some patients. Patients with DDPP low and immunotolerant normal tissue had the shortest DFS (P = 2.12E-11).

CONCLUSION

TNM stage, DDPP score, and immune competence status of normal lung are independent prognostic factors in multivariate analysis. Our findings open new avenues for prospective prognostic assessment and treatment assignment on the basis of transcriptomic profiling of tumor and normal lung tissue in patients with NSCLC.

A Novel Hybrid Deep Learning Model for Metastatic Cancer Detection

Cancer has been found as a heterogeneous disease with various subtypes and aims to destroy the body's normal cells abruptly. As a result, it is essential to detect and prognosis the distinct type of cancer since they may help cancer survivors with treatment in the early stage. It must also divide cancer patients into high- and low-risk groups. While realizing efficient detection of cancer is frequently a time-taking and exhausting task with the high possibility of pathologist errors and previous studies employed data mining and machine learning (ML) techniques to identify cancer, these strategies rely on handcrafted feature extraction techniques that result in incorrect classification. On the contrary, deep learning (DL) is robust in feature extraction and has recently been widely used for classification and detection purposes. This research implemented a novel hybrid AlexNet-gated recurrent unit (AlexNet-GRU) model for the lymph node (LN) breast cancer detection and classification. We have used a well-known Kaggle (PCam) data set to classify LN cancer samples. This study is tested and compared among three models: convolutional neural network GRU (CNN-GRU), CNN long short-term memory (CNN-LSTM), and the proposed AlexNet-GRU. The experimental results indicated that the performance metrics accuracy, precision, sensitivity, and specificity (99.50%, 98.10%, 98.90%, and 97.50) of the proposed model can reduce the pathologist errors that occur during the diagnosis process of incorrect classification and significantly better performance than CNN-GRU and CNN-LSTM models. The proposed model is compared with other recent ML/DL algorithms to analyze the model's efficiency, which reveals that the proposed AlexNet-GRU model is computationally efficient. Also, the proposed model presents its superiority over state-of-the-art methods for LN breast cancer detection and classification.

Comparative analysis of molecular signatures reveals a hybrid approach in breast cancer: Combining the Nottingham Prognostic Index with gene expressions into a hybrid signature

The diagnosis of breast cancer—including determination of prognosis and prediction—has been traditionally based on clinical and pathological characteristics such as tumor size, nodal status, and tumor grade. The decision-making process has been expanded by the recent introduction of molecular signatures. These signatures, however, have not reached the highest levels of evidence thus far. Yet they have been brought to clinical practice based on statistical significance in prospective as well as retrospective studies. Intriguingly, it has also been reported that most random sets of genes are significantly associated with disease outcome. These facts raise two highly relevant questions: What information gain do these signatures procure? How can one find a signature that is substantially better than a random set of genes? Our study addresses these questions. To address the latter question, we present a hybrid signature that joins the traditional approach with the molecular one by combining the Nottingham Prognostic Index with gene expressions in a data-driven fashion. To address the issue of information gain, we perform careful statistical analysis and comparison of the hybrid signature, gene expression lists of two commercially available tests as well as signatures selected at random, and introduce the Signature Skill Score—a simple measure to assess improvement on random signatures. Despite being based on in silico data, our research is designed to be useful for the decision-making process of oncologists and strongly supports association of random signatures with outcome. Although our study shows that none of these signatures can be considered as the main candidate for providing prognostic information, it also demonstrates that both the hybrid signature and the gene expression list of the OncotypeDx signature identify patients who may not require adjuvant chemotherapy. More importantly, we show that combining signatures substantially improves the identification of patients who do not need adjuvant chemotherapy.

Artificial intelligence, molecular subtyping, biomarkers, and precision oncology

A targeted cancer therapy is only useful if there is a way to accurately identify the tumors that are susceptible to that therapy. Thus rapid expansion in the number of available targeted cancer treatments has been accompanied by a robust effort to subdivide the traditional histological and anatomical tumor classifications into molecularly defined subtypes. This review highlights the history of the paired evolution of targeted therapies and biomarkers, reviews currently used methods for subtype identification, and discusses challenges to the implementation of precision oncology as well as possible solutions.

An Immune Signature Robustly Predicts Clinical Deterioration for Hepatitis C Virus-Related Early-Stage Cirrhosis Patients

Hepatitis C virus (HCV)-related cirrhosis leads to a heavy global burden of disease. Clinical risk stratification in HCV-related compensated cirrhosis remains a major challenge. Here, we aim to develop a signature comprised of immune-related genes to identify patients at high risk of progression and systematically analyze immune infiltration in HCV-related early-stage cirrhosis patients. Bioinformatics analysis was applied to identify immune-related genes and construct a prognostic signature in microarray data set. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) functional enrichment analyses were conducted with the “clusterProfiler” R package. Besides, the single sample gene set enrichment analysis (ssGSEA) was used to quantify immune-related risk term abundance. The nomogram and calibrate were set up via the integration of the risk score and clinicopathological characteristics to assess the effectiveness of the prognostic signature. Finally, three genes were identified and were adopted to build an immune-related prognostic signature for HCV-related cirrhosis patients. The signature was proved to be an independent risk element for HCV-related cirrhosis patients. In addition, according to the time-dependent receiver operating characteristic (ROC) curves, nomogram, and calibration plot, the prognostic model could precisely forecast the survival rate at the first, fifth, and tenth year. Notably, functional enrichment analyses indicated that cytokine activity, chemokine activity, leukocyte migration and chemotaxis, chemokine signaling pathway and viral protein interaction with cytokine and cytokine receptor were involved in HCV-related cirrhosis progression. Moreover, ssGSEA analyses revealed fierce immune-inflammatory response mechanisms in HCV progress. Generally, our work developed a robust prognostic signature that can accurately predict the overall survival, Child-Pugh class progression, hepatic decompensation, and hepatocellular carcinoma (HCC) for HCV-related early-stage cirrhosis patients. Functional enrichment and further immune infiltration analyses systematically elucidated potential immune response mechanisms.

Digital Display Precision Predictor: the prototype of a global biomarker model to guide treatments with targeted therapy and predict progression-free survival

The expanding targeted therapy landscape requires combinatorial biomarkers for patient stratification and treatment selection. This requires simultaneous exploration of multiple genes of relevant networks to account for the complexity of mechanisms that govern drug sensitivity and predict clinical outcomes. We present the algorithm, Digital Display Precision Predictor (DDPP), aiming to identify transcriptomic predictors of treatment outcome. For example, 17 and 13 key genes were derived from the literature by their association with MTOR and angiogenesis pathways, respectively, and their expression in tumor versus normal tissues was associated with the progression-free survival (PFS) of patients treated with everolimus or axitinib (respectively) using DDPP. A specific eight-gene set best correlated with PFS in six patients treated with everolimus: AKT2, TSC1, FKB-12, TSC2, RPTOR, RHEB, PIK3CA, and PIK3CB (r = 0.99, p = 5.67E-05). A two-gene set best correlated with PFS in five patients treated with axitinib: KIT and KITLG (r = 0.99, p = 4.68E-04). Leave-one-out experiments demonstrated significant concordance between observed and DDPP-predicted PFS (r = 0.9, p = 0.015) for patients treated with everolimus. Notwithstanding the small cohort and pending further prospective validation, the prototype of DDPP offers the potential to transform patients' treatment selection with a tumor- and treatment-agnostic predictor of outcomes (duration of PFS).

Breath analysis based early gastric cancer classification from deep stacked sparse autoencoder neural network

Deep learning is an emerging tool, which is regularly used for disease diagnosis in the medical field. A new research direction has been developed for the detection of early-stage gastric cancer. The computer-aided diagnosis (CAD) systems reduce the mortality rate due to their effectiveness. In this study, we proposed a new method for feature extraction using a stacked sparse autoencoder to extract the discriminative features from the unlabeled data of breath samples. A Softmax classifier was then integrated to the proposed method of feature extraction, to classify gastric cancer from the breath samples. Precisely, we identified fifty peaks in each spectrum to distinguish the EGC, AGC, and healthy persons. This CAD system reduces the distance between the input and output by learning the features and preserve the structure of the input data set of breath samples. The features were extracted from the unlabeled data of the breath samples. After the completion of unsupervised training, autoencoders with Softmax classifier were cascaded to develop a deep stacked sparse autoencoder neural network. In last, fine-tuning of the developed neural network was carried out with labeled training data to make the model more reliable and repeatable. The proposed deep stacked sparse autoencoder neural network architecture exhibits excellent results, with an overall accuracy of 98.7% for advanced gastric cancer classification and 97.3% for early gastric cancer detection using breath analysis. Moreover, the developed model produces an excellent result for recall, precision, and f score value, making it suitable for clinical application.

Cancer diagnosis with DNA molecular computation

Early and precise cancer diagnosis substantially improves patient survival. Recent work has revealed that the levels of multiple microRNAs in serum are informative as biomarkers for the diagnosis of cancers. Here, we designed a DNA molecular computation platform for the analysis of miRNA profiles in clinical serum samples. A computational classifier is first trained in silico using miRNA profiles from The Cancer Genome Atlas. This is followed by a computationally powerful but simple molecular implementation scheme using DNA, as well as an effective in situ amplification and transformation method for miRNA enrichment in serum without perturbing the original variety and quantity information. We successfully achieved rapid and accurate cancer diagnosis using clinical serum samples from 22 healthy people (8) and people with lung cancer (14) with an accuracy of 86.4%. We envision that this DNA computational platform will inspire more clinical applications towards inexpensive, non-invasive and rapid disease screening, classification and progress monitoring.

Bioinformatics profiling identifies seven immune-related risk signatures for hepatocellular carcinoma

Background

Density of tumor infiltrating lymphocytes (TIL) and expressions of certain immune-related genes have prognostic and predictive values in hepatocellular carcinoma (HCC); however, factors determining the immunophenotype of HCC patients are still unclear. In the current study, the transcript sequencing data of liver cancer were systematically analyzed to determine an immune gene marker for the prediction of clinical outcome of HCC.

Methods

RNASeq data and clinical follow-up information were downloaded from The Cancer Genome Atlas (TCGA), and the samples were assigned into high-stage and low-stage groups. Immune pathway-related genes were screened from the Molecular Signatures Database v4.0 (MsigDB) database. LASSO regression analysis was performed to identify robust immune-related biomarkers in predicting HCC clinical outcomes. Moreover, an immune gene-related prognostic model was established and validated by test sets and Gene Expression Omnibus (GEO) external validation sets.

Results

We obtained 319 immune genes from MsigDB, and the genes have different expression profiles in high-stage and low-stage of HCC. Univariate survival analysis found that 17 genes had a significant effect on HCC prognosis, among them, 13 (76.5%) genes were prognostically protective factors. Further lasso regression analysis identified seven potential prognostic markers (IL27, CD1D, NCOA6, CTSE, FCGRT, CFHR1, and APOA2) of robustness, most of which are related to tumor development. Cox regression analysis was further performed to establish a seven immune gene signature, which could stratify the risk of samples in training set, test set and external verification set (p < 0.01), and the AUC in both training set and test set was greater than 0.85, which also greater compared with previous studies.

Conclusion

This study constructed a 7-immunogenic marker as novel prognostic markers for predicting survival of HCC patients.

Identification and transfer of spatial transcriptomics signatures for cancer diagnosis

BACKGROUND

Distinguishing ductal carcinoma in situ (DCIS) from invasive ductal carcinoma (IDC) regions in clinical biopsies constitutes a diagnostic challenge. Spatial transcriptomics (ST) is an in situ capturing method, which allows quantification and visualization of transcriptomes in individual tissue sections. In the past, studies have shown that breast cancer samples can be used to study their transcriptomes with spatial resolution in individual tissue sections. Previously, supervised machine learning methods were used in clinical studies to predict the clinical outcomes for cancer types.

METHODS

We used four publicly available ST breast cancer datasets from breast tissue sections annotated by pathologists as non-malignant, DCIS, or IDC. We trained and tested a machine learning method (support vector machine) based on the expert annotation as well as based on automatic selection of cell types by their transcriptome profiles.

RESULTS

We identified expression signatures for expert annotated regions (non-malignant, DCIS, and IDC) and build machine learning models. Classification results for 798 expression signature transcripts showed high coincidence with the expert pathologist annotation for DCIS (100%) and IDC (96%). Extending our analysis to include all 25,179 expressed transcripts resulted in an accuracy of 99% for DCIS and 98% for IDC. Further, classification based on an automatically identified expression signature covering all ST spots of tissue sections resulted in prediction accuracy of 95% for DCIS and 91% for IDC.

CONCLUSIONS

This concept study suggest that the ST signatures learned from expert selected breast cancer tissue sections can be used to identify breast cancer regions in whole tissue sections including regions not trained on. Furthermore, the identified expression signatures can classify cancer regions in tissue sections not used for training with high accuracy. Expert-generated but even automatically generated cancer signatures from ST data might be able to classify breast cancer regions and provide clinical decision support for pathologists in the future.

Machine learning applications in drug development

Due to the huge amount of biological and medical data available today, along with well-established machine learning algorithms, the design of largely automated drug development pipelines can now be envisioned. These pipelines may guide, or speed up, drug discovery; provide a better understanding of diseases and associated biological phenomena; help planning preclinical wet-lab experiments, and even future clinical trials. This automation of the drug development process might be key to the current issue of low productivity rate that pharmaceutical companies currently face. In this survey, we will particularly focus on two classes of methods: sequential learning and recommender systems, which are active biomedical fields of research.

Applications of machine learning in drug discovery and development

Drug discovery and development pipelines are long, complex and depend on numerous factors. Machine learning (ML) approaches provide a set of tools that can improve discovery and decision making for well-specified questions with abundant, high-quality data. Opportunities to apply ML occur in all stages of drug discovery. Examples include target validation, identification of prognostic biomarkers and analysis of digital pathology data in clinical trials. Applications have ranged in context and methodology, with some approaches yielding accurate predictions and insights. The challenges of applying ML lie primarily with the lack of interpretability and repeatability of ML-generated results, which may limit their application. In all areas, systematic and comprehensive high-dimensional data still need to be generated. With ongoing efforts to tackle these issues, as well as increasing awareness of the factors needed to validate ML approaches, the application of ML can promote data-driven decision making and has the potential to speed up the process and reduce failure rates in drug discovery and development.

Novel Model for Comprehensive Assessment of Robust Prognostic Gene Signature in Ovarian Cancer Across Different Independent Datasets

Different analytical methods or models can often find completely different prognostic biomarkers for the same cancer. In the study of prognostic molecular biomarkers of ovarian cancer (OvCa), different studies have reported a variety of prognostic gene signatures. In the current study, based on geometric concepts, the linearity-clustering phase diagram with integrated P-value (LCP) method was used to comprehensively consider three indicators that are commonly employed to estimate the quality of a prognostic gene signature model. The three indicators, namely, concordance index, area under the curve, and level of the hazard ratio were determined via calculation of the prognostic index of various gene signatures from different datasets. As evaluation objects, we selected 13 gene signature models (Cox regression model) and 16 OvCa genomic datasets (including gene expression information and follow-up data) from published studies. The results of LCP showed that three models were universal and better than other models. In addition, combining the three models into one model showed the best performance in all datasets by LCP calculation. The combination gene signature model provides a more reliable model and could be validated in various datasets of OvCa. Thus, our method and findings can provide more accurate prognostic biomarkers and effective reference for the precise clinical treatment of OvCa.

Cross comparison and prognostic assessment of breast cancer multigene signatures in a large population-based contemporary clinical series

Multigene expression signatures provide a molecular subdivision of early breast cancer associated with patient outcome. A gap remains in the validation of such signatures in clinical treatment groups of patients within population-based cohorts of unselected primary breast cancer representing contemporary disease stages and current treatments. A cohort of 3520 resectable breast cancers with RNA sequencing data included in the population-based SCAN-B initiative (ClinicalTrials.gov ID NCT02306096) were selected from a healthcare background population of 8587 patients diagnosed within the years 2010-2015. RNA profiles were classified according to 19 reported gene signatures including both gene expression subtypes (e.g. PAM50, IC10, CIT) and risk predictors (e.g. Oncotype DX, 70-gene, ROR). Classifications were analyzed in nine adjuvant clinical assessment groups: TNBC-ACT (adjuvant chemotherapy, n = 239), TNBC-untreated (n = 82), HER2+/ER- with anti-HER2+ ACT treatment (n = 110), HER2+/ER+ with anti-HER2 + ACT + endocrine treatment (n = 239), ER+/HER2-/LN- with endocrine treatment (n = 1113), ER+/HER2-/LN- with endocrine + ACT treatment (n = 243), ER+/HER2-/LN+ with endocrine treatment (n = 423), ER+/HER2-/LN+ with endocrine + ACT treatment (n = 433), and ER+/HER2-/LN- untreated (n = 200). Gene signature classification (e.g., proportion low-, high-risk) was generally well aligned with stratification based on current immunohistochemistry-based clinical practice. Most signatures did not provide any further risk stratification in TNBC and HER2+/ER- disease. Risk classifier agreement (low-, medium/intermediate-, high-risk groups) in ER+ assessment groups was on average 50-60% with occasional pair-wise comparisons having <30% agreement. Disregarding the intermediate-risk groups, the exact agreement between low- and high-risk groups was on average ~80-95%, for risk prediction signatures across all assessment groups. Outcome analyses were restricted to assessment groups of TNBC-ACT and endocrine treated ER+/HER2-/LN- and ER+/HER2-/LN+ cases. For ER+/HER2- disease, gene signatures appear to contribute additional prognostic value even at a relatively short follow-up time. Less apparent prognostic value was observed in the other groups for the tested signatures. The current study supports the usage of gene expression signatures in specific clinical treatment groups within population-based breast cancer. It also stresses the need of further development to reach higher consensus in individual patient classifications, especially for intermediate-risk patients, and the targeting of patients where current gene signatures and prognostic variables provide little support in clinical decision-making.

Genomic and transcriptomic profiling expands precision cancer medicine: the WINTHER trial

Precision medicine focuses on DNA abnormalities, but not all tumors have tractable genomic alterations. The WINTHER trial ( NCT01856296 ) navigated patients to therapy on the basis of fresh biopsy-derived DNA sequencing (arm A; 236 gene panel) or RNA expression (arm B; comparing tumor to normal). The clinical management committee (investigators from five countries) recommended therapies, prioritizing genomic matches; physicians determined the therapy given. Matching scores were calculated post-hoc for each patient, according to drugs received: for DNA, the number of alterations matched divided by the total alteration number; for RNA, expression-matched drug ranks. Overall, 303 patients consented; 107 (35%; 69 in arm A and 38 in arm B) were evaluable for therapy. The median number of previous therapies was three. The most common diagnoses were colon, head and neck, and lung cancers. Among the 107 patients, the rate of stable disease ≥6 months and partial or complete response was 26.2% (arm A: 23.2%; arm B: 31.6% (P = 0.37)). The patient proportion with WINTHER versus previous therapy progression-free survival ratio of >1.5 was 22.4%, which did not meet the pre-specified primary end point. Fewer previous therapies, better performance status and higher matching score correlated with longer progression-free survival (all P < 0.05, multivariate). Our study shows that genomic and transcriptomic profiling are both useful for improving therapy recommendations and patient outcome, and expands personalized cancer treatment.

GradientScanSurv-An exhaustive association test method for gene expression data with censored survival outcome

Accurate assessment of the association between continuous variables such as gene expression and survival is a critical aspect of precision medicine. In this report, we provide a review of some of the available survival analysis and validation tools by referencing published studies that have utilized these tools. We have identified pitfalls associated with the assumptions inherent in those applications that have the potential to impact scientific research through their potential bias. In order to overcome these pitfalls, we have developed a novel method that enables the logrank test method to handle continuous variables that comprehensively evaluates survival association with derived aggregate statistics. This is accomplished by exhaustively considering all the cutpoints across the full expression gradient. Direct side-by-side comparisons, global ROC analysis, and evaluation of the ability to capture relevant biological themes based on current understanding of RAS biology all demonstrated that the new method shows better consistency between multiple datasets of the same disease, better reproducibility and robustness, and better detection power to uncover biological relevance within the selected datasets over the available survival analysis methods on univariate gene expression and penalized linear model-based methods.

The germline genetic component of drug sensitivity in cancer cell lines

Patients with seemingly the same tumour can respond very differently to treatment. There are strong, well-established effects of somatic mutations on drug efficacy, but there is at-most anecdotal evidence of a germline component to drug response. Here, we report a systematic survey of how inherited germline variants affect drug susceptibility in cancer cell lines. We develop a joint analysis approach that leverages both germline and somatic variants, before applying it to screening data from 993 cell lines and 265 drugs. Surprisingly, we find that the germline contribution to variation in drug susceptibility can be as large or larger than effects due to somatic mutations. Several of the associations identified have a direct relationship to the drug target. Finally, using 17-AAG response as an example, we show how germline effects in combination with transcriptomic data can be leveraged for improved patient stratification and to identify new markers for drug sensitivity.

Little is known about the contribution of germline genetic variants to cancer drug sensitivity. Here, the authors devise an approach for joint analysis of germline variants and somatic mutations, identifying substantial germline contributions to variation in drug sensitivity.

A molecular multi-gene classifier for disease diagnostics

Despite its early promise as a diagnostic and prognostic tool, gene expression profiling remains cost-prohibitive and challenging to implement in a clinical setting. Here, we introduce a molecular computation strategy for analysing the information contained in complex gene expression signatures without the need for costly instrumentation. Our workflow begins by training a computational classifier on labelled gene expression data. This in silico classifier is then realized at the molecular level to enable expression analysis and classification of previously uncharacterized samples. Classification occurs through a series of molecular interactions between RNA inputs and engineered DNA probes designed to differentially weigh each input according to its importance. We validate our technology with two applications: a classifier for early cancer diagnostics and a classifier for differentiating viral and bacterial respiratory infections based on host gene expression. Together, our results demonstrate a general and modular framework for low-cost gene expression analysis.

A probabilistic pathway score (PROPS) for classification with applications to inflammatory bowel disease

Summary

Gene-based supervised machine learning classification models have been widely used to differentiate disease states, predict disease progression and determine effective treatment options. However, many of these classifiers are sensitive to noise and frequently do not replicate in external validation sets. For complex, heterogeneous diseases, these classifiers are further limited by being unable to capture varying combinations of genes that lead to the same phenotype. Pathway-based classification can overcome these challenges by using robust, aggregate features to represent biological mechanisms. In this work, we developed a novel pathway-based approach, PRObabilistic Pathway Score, which uses genes to calculate individualized pathway scores for classification. Unlike previous individualized pathway-based classification methods that use gene sets, we incorporate gene interactions using probabilistic graphical models to more accurately represent the underlying biology and achieve better performance. We apply our method to differentiate two similar complex diseases, ulcerative colitis (UC) and Crohn's disease (CD), which are the two main types of inflammatory bowel disease (IBD). Using five IBD datasets, we compare our method against four gene-based and four alternative pathway-based classifiers in distinguishing CD from UC. We demonstrate superior classification performance and provide biological insight into the top pathways separating CD from UC.

Availability and Implementation

PROPS is available as a R package, which can be downloaded at http://simtk.org/home/props or on Bioconductor.

Contact

rbaltman@stanford.edu.

Supplementary information

Supplementary data are available at Bioinformatics online.

Immune signature profiling identified predictive and prognostic factors for esophageal squamous cell carcinoma

Understanding interactions between tumor and the host immune system holds great promise to uncover biomarkers for targeted therapies and clinical outcomes. However, systematical analysis of immune signatures in esophageal squamous cell carcinoma (ESCC) remains largely unstudied. In this study, immune signatures containing 708 immune related genes were curated from mRNA microarray data with tumor and paired normal tissues from 119 ESCC patients. Differential expression and survival analysis were performed with validations from Human Protein Atlas and an independent cohort of 110 ESCC patients by immunohistochemistry staining. We identified a total of 186 significantly dysregulated genes in ESCC, including downregulated genes SPINK5, IL1RN and upregulated genes SPP1 and PLAU, which were further confirmed in Human Protein Atlas data. Moreover, nine immune related genes (ABL1, ATF2, ATG5, C6, CD38, HMGB1, ICOSLG, IL12RB2 and PLAU) were significantly associated with patients' overall survival, among which, prognostic model was built including three independent factors ABL1, CD38 and ICOSLG. Validation by immunohistochemistry staining suggested that combination with tumor infiltrated CD4+ and CD8+ T lymphocytes would yield higher performance in distinguishing cases as high or low risk of unfavorable prognosis. In summary, we profiled the immune status in ESCC and established predictive and prognostic factors for ESCC, which could reflect immune disorders within tumor microenvironments and independently distinguish patients with a high risk of reduced survival, providing novel predictive and therapeutic targets for ESCC patients in the future.

A 16 Yin Yang gene expression ratio signature for ER+/node- breast cancer

Breast cancer is one of the leading causes of cancer death in women. It is a complex and heterogeneous disease with different clinical outcomes. Stratifying patients into subgroups with different outcomes could help guide clinical decision making. In this study, we used two opposing groups of genes, Yin and Yang, to develop a prognostic expression ratio signature. Using the METABRIC cohort we identified a16-gene signature capable of stratifying breast cancer patients into four risk levels with intention that low-risk patients would not undergo adjuvant systemic therapy, intermediate-low-risk patients will be treated with hormonal therapy only, and intermediate-high- and high-risk groups will be treated by chemotherapy in addition to the hormonal therapy. The 16-gene signature for four risk level stratifications of breast cancer patients has been validated using 14 independent datasets. Notably, the low-risk group (n = 51) of 205 estrogen receptor-positive and node negative (ER+/node-) patients from three different datasets who had not had any systemic adjuvant therapy had 100% 15-year disease-specific survival rate. The Concordance Index of YMR for ER+/node negative patients is close to the commercially available signatures. However, YMR showed more significance (HR = 3.7, p = 8.7e-12) in stratifying ER+/node- subgroup than OncotypeDx (HR = 2.7, p = 1.3e-7), MammaPrint (HR = 2.5, p = 5.8e-7), rorS (HR = 2.4, p = 1.4e-6), and NPI (HR = 2.6, p = 1.2e-6). YMR signature may be developed as a clinical tool to select a subgroup of low-risk ER+/node- patients who do not require any adjuvant hormonal therapy (AHT).

Distance in cancer gene expression from stem cells predicts patient survival

The degree of histologic cellular differentiation of a cancer has been associated with prognosis but is subjectively assessed. We hypothesized that information about tumor differentiation of individual cancers could be derived objectively from cancer gene expression data, and would allow creation of a cancer phylogenetic framework that would correlate with clinical, histologic and molecular characteristics of the cancers, as well as predict prognosis. Here we utilized mRNA expression data from 4,413 patient samples with 7 diverse cancer histologies to explore the utility of ordering samples by their distance in gene expression from that of stem cells. A differentiation baseline was obtained by including expression data of human embryonic stem cells (hESC) and human mesenchymal stem cells (hMSC) for solid tumors, and of hESC and CD34+ cells for liquid tumors. We found that the correlation distance (the degree of similarity) between the gene expression profile of a tumor sample and that of stem cells orients cancers in a clinically coherent fashion. For all histologies analyzed (including carcinomas, sarcomas, and hematologic malignancies), patients with cancers with gene expression patterns most similar to that of stem cells had poorer overall survival. We also found that the genes in all undifferentiated cancers of diverse histologies that were most differentially expressed were associated with up-regulation of specific oncogenes and down-regulation of specific tumor suppressor genes. Thus, a stem cell-oriented phylogeny of cancers allows for the derivation of a novel cancer gene expression signature found in all undifferentiated forms of diverse cancer histologies, that is competitive in predicting overall survival in cancer patients compared to previously published prediction models, and is coherent in that gene expression was associated with up-regulation of specific oncogenes and down-regulation of specific tumor suppressor genes associated with regulation of the multicellular state.

Characteristics and Validation Techniques for PCA-Based Gene-Expression Signatures

Background. Many gene-expression signatures exist for describing the biological state of profiled tumors. Principal Component Analysis (PCA) can be used to summarize a gene signature into a single score. Our hypothesis is that gene signatures can be validated when applied to new datasets, using inherent properties of PCA. Results. This validation is based on four key concepts. Coherence: elements of a gene signature should be correlated beyond chance. Uniqueness: the general direction of the data being examined can drive most of the observed signal. Robustness: if a gene signature is designed to measure a single biological effect, then this signal should be sufficiently strong and distinct compared to other signals within the signature. Transferability: the derived PCA gene signature score should describe the same biology in the target dataset as it does in the training dataset. Conclusions. The proposed validation procedure ensures that PCA-based gene signatures perform as expected when applied to datasets other than those that the signatures were trained upon. Complex signatures, describing multiple independent biological components, are also easily identified.

Distinct splicing signatures affect converged pathways in myelodysplastic syndrome patients carrying mutations in different splicing regulators

Myelodysplastic syndromes (MDS) are heterogeneous myeloid disorders with prevalent mutations in several splicing factors, but the splicing programs linked to specific mutations or MDS in general remain to be systematically defined. We applied RASL-seq, a sensitive and cost-effective platform, to interrogate 5502 annotated splicing events in 169 samples from MDS patients or healthy individuals. We found that splicing signatures associated with normal hematopoietic lineages are largely related to cell signaling and differentiation programs, whereas MDS-linked signatures are primarily involved in cell cycle control and DNA damage responses. Despite the shared roles of affected splicing factors in the 3' splice site definition, mutations in U2AF1, SRSF2, and SF3B1 affect divergent splicing programs, and interestingly, the affected genes fall into converging cancer-related pathways. A risk score derived from 11 splicing events appears to be independently associated with an MDS prognosis and AML transformation, suggesting potential clinical relevance of altered splicing patterns in MDS.

Statistical controversies in clinical research: prognostic gene signatures are not (yet) useful in clinical practice

With the genomic revolution and the era of targeted therapy, prognostic and predictive gene signatures are becoming increasingly important in clinical research. They are expected to assist prognosis assessment and therapeutic decision making. Notwithstanding, an evidence-based approach is needed to bring gene signatures from the laboratory to clinical practice. In early breast cancer, multiple prognostic gene signatures are commercially available without having formally reached the highest levels of evidence-based criteria. We discuss specific concepts for developing and validating a prognostic signature and illustrate them with contemporary examples in breast cancer. When a prognostic signature has not been developed for predicting the magnitude of relative treatment benefit through an interaction effect, it may be wishful thinking to test its predictive value. We propose that new gene signatures be built specifically for predicting treatment effects for future patients and outline an approach for this using a cross-validation scheme in a standard phase III trial. Replication in an independent trial remains essential.

A simplified interventional mapping system (SIMS) for the selection of combinations of targeted treatments in non-small cell lung cancer

Non-small cell lung cancer (NSCLC) is a leading cause of death worldwide. Targeted monotherapies produce high regression rates, albeit for limited patient subgroups, who inevitably succumb. We present a novel strategy for identifying customized combinations of triplets of targeted agents, utilizing a simplified interventional mapping system (SIMS) that merges knowledge about existent drugs and their impact on the hallmarks of cancer. Based on interrogation of matched lung tumor and normal tissue using targeted genomic sequencing, copy number variation, transcriptomics, and miRNA expression, the activation status of 24 interventional nodes was elucidated. An algorithm was developed to create a scoring system that enables ranking of the activated interventional nodes for each patient. Based on the trends of co-activation at interventional points, combinations of drug triplets were defined in order to overcome resistance. This methodology will inform a prospective trial to be conducted by the WIN consortium, aiming to significantly impact survival in metastatic NSCLC and other malignancies.

SNRFCB: sub-network based random forest classifier for predicting chemotherapy benefit on survival for cancer treatment

Adjuvant chemotherapy (CTX) should be individualized to provide potential survival benefit and avoid potential harm to cancer patients. Our goal was to establish a computational approach for making personalized estimates of the survival benefit from adjuvant CTX. We developed Sub-Network based Random Forest classifier for predicting Chemotherapy Benefit (SNRFCB) based gene expression datasets of lung cancer. The SNRFCB approach was then validated in independent test cohorts for identifying chemotherapy responder cohorts and chemotherapy non-responder cohorts. SNRFCB involved the pre-selection of gene sub-network signatures based on the mutations and on protein-protein interaction data as well as the application of the random forest algorithm to gene expression datasets. Adjuvant CTX was significantly associated with the prolonged overall survival of lung cancer patients in the chemotherapy responder group (P = 0.008), but it was not beneficial to patients in the chemotherapy non-responder group (P = 0.657). Adjuvant CTX was significantly associated with the prolonged overall survival of lung cancer squamous cell carcinoma (SQCC) subtype patients in the chemotherapy responder cohorts (P = 0.024), but it was not beneficial to patients in the chemotherapy non-responder cohorts (P = 0.383). SNRFCB improved prediction performance as compared to the machine learning method, support vector machine (SVM). To test the general applicability of the predictive model, we further applied the SNRFCB approach to human breast cancer datasets and also observed superior performance. SNRFCB could provide recurrent probability for individual patients and identify which patients may benefit from adjuvant CTX in clinical trials.

Clinicopathological indices to predict hepatocellular carcinoma molecular classification

BACKGROUND & AIMS

Hepatocellular carcinoma (HCC) is the second most lethal cancer caused by lack of effective therapies. Although promising, HCC molecular classification, which enriches potential responders to specific therapies, has not yet been assessed in clinical trials of anti-HCC drugs. We aimed to overcome these challenges by developing clinicopathological surrogate indices of HCC molecular classification.

METHODS

Hepatocellular carcinoma classification defined in our previous transcriptome meta-analysis (S1, S2 and S3 subclasses) was implemented in an FDA-approved diagnostic platform (Elements assay, NanoString). Ninety-six HCC tumours (training set) were assayed to develop molecular subclass-predictive indices based on clinicopathological features, which were independently validated in 99 HCC tumours (validation set). Molecular deregulations associated with the histopathological features were determined by pathway analysis. Sample sizes for HCC clinical trials enriched with specific molecular subclasses were determined.

RESULTS

Hepatocellular carcinoma subclass-predictive indices were steatohepatitic (SH)-HCC variant and immune cell infiltrate for S1 subclass, macrotrabecular/compact pattern, lack of pseudoglandular pattern, and high serum alpha-foetoprotein (>400 ng/ml) for S2 subclass, and microtrabecular pattern, lack of SH-HCC and clear cell variants, and lower histological grade for S3 subclass. Macrotrabecular/compact pattern, a predictor of S2 subclass, was associated with the activation of therapeutically targetable oncogene YAP and stemness markers EPCAM/KRT19. BMP4 was associated with pseudoglandular pattern. Subclass-predictive indices-based patient enrichment reduced clinical trial sample sizes from 121, 184 and 53 to 30, 43 and 22 for S1, S2 and S3 subclass-targeting therapies respectively.

CONCLUSIONS

Hepatocellular carcinoma molecular subclasses can be enriched by clinicopathological indices tightly associated with deregulation of therapeutically targetable molecular pathways.

A genomic and clinical prognostic index for hepatitis C-related early-stage cirrhosis that predicts clinical deterioration

OBJECTIVE

The number of patients with HCV-related cirrhosis is increasing, leading to a rising risk of complications and death. Prognostic stratification in patients with early-stage cirrhosis is still challenging. We aimed to develop and validate a clinically useful prognostic index based on genomic and clinical variables to identify patients at high risk of disease progression.

DESIGN

We developed a prognostic index, comprised of a 186-gene signature validated in our previous genome-wide profiling study, bilirubin (>1 mg/dL) and platelet count (<100,000/mm(3)), in an Italian HCV cirrhosis cohort (training cohort, n=216, median follow-up 10 years). The gene signature test was implemented using a digital transcript counting (nCounter) assay specifically developed for clinical use and the prognostic index was evaluated using archived specimens from an independent cohort of HCV-related cirrhosis in the USA (validation cohort, n=145, median follow-up 8 years).

RESULTS

In the training cohort, the prognostic index was associated with hepatic decompensation (HR=2.71, p=0.003), overall death (HR=6.00, p<0.001), hepatocellular carcinoma (HR=3.31, p=0.001) and progression of Child-Turcotte-Pugh class (HR=6.70, p<0.001). The patients in the validation cohort were stratified into high-risk (16%), intermediate-risk (42%) or low-risk (42%) groups by the prognostic index. The high-risk group had a significantly increased risk of hepatic decompensation (HR=7.36, p<0.001), overall death (HR=3.57, p=0.002), liver-related death (HR=6.49, p<0.001) and all liver-related adverse events (HR=4.98, p<0.001).

CONCLUSIONS

A genomic and clinical prognostic index readily available for clinical use was successfully validated, warranting further clinical evaluation for prognostic prediction and clinical trial stratification and enrichment for preventive interventions.

The prognostic landscape of genes and infiltrating immune cells across human cancers

Molecular profiles of tumors and tumor-associated cells hold great promise as biomarkers of clinical outcomes. However, existing data sets are fragmented and difficult to analyze systematically. Here we present a pan-cancer resource and meta-analysis of expression signatures from ∼18,000 human tumors with overall survival outcomes across 39 malignancies. By using this resource, we identified a forkhead box MI (FOXM1) regulatory network as a major predictor of adverse outcomes, and we found that expression of favorably prognostic genes, including KLRB1 (encoding CD161), largely reflect tumor-associated leukocytes. By applying CIBERSORT, a computational approach for inferring leukocyte representation in bulk tumor transcriptomes, we identified complex associations between 22 distinct leukocyte subsets and cancer survival. For example, tumor-associated neutrophil and plasma cell signatures emerged as significant but opposite predictors of survival for diverse solid tumors, including breast and lung adenocarcinomas. This resource and associated analytical tools (http://precog.stanford.edu) may help delineate prognostic genes and leukocyte subsets within and across cancers, shed light on the impact of tumor heterogeneity on cancer outcomes, and facilitate the discovery of biomarkers and therapeutic targets.

Gene Expression Signatures Based on Variability can Robustly Predict Tumor Progression and Prognosis

Gene expression signatures are commonly used to create cancer prognosis and diagnosis methods, yet only a small number of them are successfully deployed in the clinic since many fail to replicate performance on subsequent validation. A primary reason for this lack of reproducibility is the fact that these signatures attempt to model the highly variable and unstable genomic behavior of cancer. Our group recently introduced gene expression anti-profiles as a robust methodology to derive gene expression signatures based on the observation that while gene expression measurements are highly heterogeneous across tumors of a specific cancer type relative to the normal tissue, their degree of deviation from normal tissue expression in specific genes involved in tissue differentiation is a stable tumor mark that is reproducible across experiments and cancer types. Here we show that constructing gene expression signatures based on variability and the anti-profile approach yields classifiers capable of successfully distinguishing benign growths from cancerous growths based on deviation from normal expression. We then show that this same approach generates stable and reproducible signatures that predict probability of relapse and survival based on tumor gene expression. These results suggest that using the anti-profile framework for the discovery of genomic signatures is an avenue leading to the development of reproducible signatures suitable for adoption in clinical settings.

Cancer biomarker discovery and validation

With the emergence of genomic profiling technologies and selective molecular targeted therapies, biomarkers play an increasingly important role in the clinical management of cancer patients. Single gene/protein or multi-gene "signature"-based assays have been introduced to measure specific molecular pathway deregulations that guide therapeutic decision-making as predictive biomarkers. Genome-based prognostic biomarkers are also available for several cancer types for potential incorporation into clinical prognostic staging systems or practice guidelines. However, there is still a large gap between initial biomarker discovery studies and their clinical translation due to the challenges in the process of cancer biomarker development. In this review we summarize the steps of biomarker development, highlight key issues in successful validation and implementation, and overview representative examples in the oncology field. We also discuss regulatory issues and future perspectives in the era of big data analysis and precision medicine.

Expression of tumor necrosis factor-alpha-mediated genes predicts recurrence-free survival in lung cancer

In this study, we conducted a meta-analysis on high-throughput gene expression data to identify TNF-α-mediated genes implicated in lung cancer. We first investigated the gene expression profiles of two independent TNF-α/TNFR KO murine models. The EGF receptor signaling pathway was the top pathway associated with genes mediated by TNF-α. After matching the TNF-α-mediated mouse genes to their human orthologs, we compared the expression patterns of the TNF-α-mediated genes in normal and tumor lung tissues obtained from humans. Based on the TNF-α-mediated genes that were dysregulated in lung tumors, we developed a prognostic gene signature that effectively predicted recurrence-free survival in lung cancer in two validation cohorts. Resampling tests suggested that the prognostic power of the gene signature was not by chance, and multivariate analysis suggested that this gene signature was independent of the traditional clinical factors and enhanced the identification of lung cancer patients at greater risk for recurrence.

Survival analysis tools in genomics research

There is an increasing demand to determine the clinical implication of experimental findings in molecular biomedical research. Survival (or failure time) analysis methodologies have been adapted to the analysis of genomics data to link molecular information with clinical outcomes of interest. Genome-wide molecular profiles have served as sources for discovery of predictive/prognostic biomarkers as well as therapeutic targets in the past decade. In this review, we overview currently available software, web applications, and databases specifically developed for survival analysis in genomics research and discuss issues in assessing clinical utility of molecular features derived from genomic profiling.

Machine learning applications in cancer prognosis and prediction

Cancer has been characterized as a heterogeneous disease consisting of many different subtypes. The early diagnosis and prognosis of a cancer type have become a necessity in cancer research, as it can facilitate the subsequent clinical management of patients. The importance of classifying cancer patients into high or low risk groups has led many research teams, from the biomedical and the bioinformatics field, to study the application of machine learning (ML) methods. Therefore, these techniques have been utilized as an aim to model the progression and treatment of cancerous conditions. In addition, the ability of ML tools to detect key features from complex datasets reveals their importance. A variety of these techniques, including Artificial Neural Networks (ANNs), Bayesian Networks (BNs), Support Vector Machines (SVMs) and Decision Trees (DTs) have been widely applied in cancer research for the development of predictive models, resulting in effective and accurate decision making. Even though it is evident that the use of ML methods can improve our understanding of cancer progression, an appropriate level of validation is needed in order for these methods to be considered in the everyday clinical practice. In this work, we present a review of recent ML approaches employed in the modeling of cancer progression. The predictive models discussed here are based on various supervised ML techniques as well as on different input features and data samples. Given the growing trend on the application of ML methods in cancer research, we present here the most recent publications that employ these techniques as an aim to model cancer risk or patient outcomes.

Cost Effectiveness of Molecular Profiling for Adjuvant Decision Making in Patients With Node-Negative Breast Cancer

Purpose

To conduct an economic evaluation of the 70-gene signature used to guide adjuvant chemotherapy decision making both in patients with node-negative breast cancer (NNBC) and in the subgroup of estrogen receptor (ER) –positive patients.

Patients and Methods

We used a mixed approach combining patient-level data from a multicenter validation study of the 70-gene signature (untreated patients) and secondary sources for chemotherapy efficacy, unit costs, and utility values. Three strategies on which to base the decision to administer adjuvant chemotherapy were compared: the 70-gene signature, Adjuvant! Online, and chemotherapy in all patients. In the base-case analysis, costs from the French National Insurance Scheme, life-years (LYs), and quality-adjusted life-years (QALYs) were computed for the three strategies over a 10-year period. Cost-effectiveness acceptability curves using the net monetary benefit were computed, combining bootstrap and probabilistic sensitivity analyses.

Results

The mean differences in LYs and QALYs were similar between the three strategies. The 70-gene signature strategy was associated with a higher cost, with a mean difference of €2,037 (range, €1,472 to €2,515) compared with Adjuvant! Online and of €657 (95% CI, −€642 to €3,130) compared with systematic chemotherapy. For a €50,000 per QALY willingness-to-pay threshold, the probability of being the most cost-effective strategy was 92% (76% in ER-positive patients) for the Adjuvant! Online strategy, 6% (4% in ER-positive patients) for the systematic chemotherapy strategy, and 2% (20% in ER-positive patients) for the 70-gene strategy.

Conclusion

Optimizing adjuvant chemotherapy decision making based on the 70-gene signature is unlikely to be cost effective in patients with NNBC.

Expression profiling of mitochondrial voltage-dependent anion channel-1 associated genes predicts recurrence-free survival in human carcinomas

Background

Mitochondrial voltage-dependent anion channels (VDACs) play a key role in mitochondria-mediated apoptosis. Both in vivo and in vitro evidences indicate that VDACs are actively involved in tumor progression. Specifically, VDAC-1, one member of the VDAC family, was thought to be a potential anti-cancer therapeutic target. Our previous study demonstrated that the human gene VDAC1 (encoding the VDAC-1 isoform) was significantly up-regulated in lung tumor tissue compared with normal tissue. Also, we found a significant positive correlation between the gene expression of VDAC1 and histological grade in breast cancer. However, the prognostic power of VDAC1 and its associated genes in human cancers is largely unknown.

Methods

We systematically analyzed the expression pattern of VDAC1 and its interacting genes in breast, colon, liver, lung, pancreatic, and thyroid cancers. The genes differentially expressed between normal and tumor tissues in human carcinomas were identified.

Results

The expression level of VDAC1 was uniformly up-regulated in tumor tissue compared with normal tissue in breast, colon, liver, lung, pancreatic, and thyroid cancers. Forty-four VDAC1 interacting genes were identified as being commonly differentially expressed between normal and tumor tissues in human carcinomas. We designated VDAC1 and the 44 dysregulated interacting genes as the VDAC1 associated gene signature (VAG). We demonstrate that the VAG signature is a robust prognostic biomarker to predict recurrence-free survival in breast, colon, and lung cancers, and is independent of standard clinical and pathological prognostic factors.

Conclusions

VAG represents a promising prognostic biomarker in human cancers, which may enhance prediction accuracy in identifying patients at higher risk for recurrence. Future therapies aimed specifically at VDAC1 associated genes may lead to novel agents in the treatment of cancer.

A program to identify prognostic and predictive gene signatures

BACKGROUND

The advent of high-throughput technologies to profile human tumors has generated unprecedented insight into our molecular understanding of cancer. However, analysis of such high dimensional data is challenging and requires significant expertise which is not routinely available to many cancer researchers.

RESULTS

To overcome this limitation, we developed a freely accessible and user friendly Program to Identify Molecular Signatures (PIMS). Importantly, such signatures allow important insight into cancer biology, as well as provide clinical tools to identify potential biomarkers that might provide means to accurately stratify patients into different risk or treatment groups. We evaluated the performance of PIMS by identifying and testing predictive and prognostic gene signatures for breast cancer, using multiple breast tumor microarray cohorts representing hundreds of patients. Importantly, PIMS identified signatures classified patients into high and low risk groups with at least similar performance to other commonly used gene signature selection techniques.

CONCLUSIONS

Our program is contained entirely within a Microsoft Excel file and therefore requires no installation of any additional programs or training. Hence, PIMS provides an accessible tool for cancer researchers to identify predictive and prognostic gene signatures to advance their research.

Functional characterization of breast cancer using pathway profiles

BACKGROUND

The molecular characteristics of human diseases are often represented by a list of genes termed "signature genes". A significant challenge facing this approach is that of reproducibility: signatures developed on a set of patients may fail to perform well on different sets of patients. As diseases are resulted from perturbed cellular functions, irrespective of the particular genes that contribute to the function, it may be more appropriate to characterize diseases based on these perturbed cellular functions.

METHODS

We proposed a profile-based approach to characterize a disease using a binary vector whose elements indicate whether a given function is perturbed based on the enrichment analysis of expression data between normal and tumor tissues. Using breast cancer and its four primary clinically relevant subtypes as examples, this approach is evaluated based on the reproducibility, accuracy and resolution of the resulting pathway profiles.

RESULTS

Pathway profiles for breast cancer and its subtypes are constructed based on data obtained from microarray and RNA-Seq data sets provided by The Cancer Genome Atlas (TCGA), and an additional microarray data set provided by The European Genome-phenome Archive (EGA). An average reproducibility of 68% is achieved between different data sets (TCGA microarray vs. EGA microarray data) and 67% average reproducibility is achieved between different technologies (TCGA microarray vs. TCGA RNA-Seq data). Among the enriched pathways, 74% of them are known to be associated with breast cancer or other cancers. About 40% of the identified pathways are enriched in all four subtypes, with 4, 2, 4, and 7 pathways enriched only in luminal A, luminal B, triple-negative, and HER2+ subtypes, respectively. Comparison of profiles between subtypes, as well as other diseases, shows that luminal A and luminal B subtypes are more similar to the HER2+ subtype than to the triple-negative subtype, and subtypes of breast cancer are more likely to be closer to each other than to other diseases.

CONCLUSIONS

Our results demonstrate that pathway profiles can successfully characterize both common and distinct functional characteristics of four subtypes of breast cancer and other related diseases, with acceptable reproducibility, high accuracy and reasonable resolution.

Risk prediction for late-stage ovarian cancer by meta-analysis of 1525 patient samples

BACKGROUND

Ovarian cancer causes more than 15000 deaths per year in the United States. The survival of patients is quite heterogeneous, and accurate prognostic tools would help with the clinical management of these patients.

METHODS

We developed and validated two gene expression signatures, the first for predicting survival in advanced-stage, serous ovarian cancer and the second for predicting debulking status. We integrated 13 publicly available datasets totaling 1525 subjects. We trained prediction models using a meta-analysis variation on the compound covariable method, tested models by a "leave-one-dataset-out" procedure, and validated models in additional independent datasets. Selected genes from the debulking signature were validated by immunohistochemistry and quantitative reverse-transcription polymerase chain reaction (qRT-PCR) in two further independent cohorts of 179 and 78 patients, respectively. All statistical tests were two-sided.

RESULTS

The survival signature stratified patients into high- and low-risk groups (hazard ratio = 2.19; 95% confidence interval [CI] = 1.84 to 2.61) statistically significantly better than the TCGA signature (P = .04). POSTN, CXCL14, FAP, NUAK1, PTCH1, and TGFBR2 were validated by qRT-PCR (P < .05) and POSTN, CXCL14, and phosphorylated Smad2/3 were validated by immunohistochemistry (P < .001) as independent predictors of debulking status. The sum of immunohistochemistry intensities for these three proteins provided a tool that classified 92.8% of samples correctly in high- and low-risk groups for suboptimal debulking (area under the curve = 0.89; 95% CI = 0.84 to 0.93).

CONCLUSIONS

Our survival signature provides the most accurate and validated prognostic model for early- and advanced-stage high-grade, serous ovarian cancer. The debulking signature accurately predicts the outcome of cytoreductive surgery, potentially allowing for stratification of patients for primary vs secondary cytoreduction.

Risk prediction for late-stage ovarian cancer by meta-analysis of 1525 patient samples

BACKGROUND

Ovarian cancer causes more than 15000 deaths per year in the United States. The survival of patients is quite heterogeneous, and accurate prognostic tools would help with the clinical management of these patients.

METHODS

We developed and validated two gene expression signatures, the first for predicting survival in advanced-stage, serous ovarian cancer and the second for predicting debulking status. We integrated 13 publicly available datasets totaling 1525 subjects. We trained prediction models using a meta-analysis variation on the compound covariable method, tested models by a "leave-one-dataset-out" procedure, and validated models in additional independent datasets. Selected genes from the debulking signature were validated by immunohistochemistry and quantitative reverse-transcription polymerase chain reaction (qRT-PCR) in two further independent cohorts of 179 and 78 patients, respectively. All statistical tests were two-sided.

RESULTS

The survival signature stratified patients into high- and low-risk groups (hazard ratio = 2.19; 95% confidence interval [CI] = 1.84 to 2.61) statistically significantly better than the TCGA signature (P = .04). POSTN, CXCL14, FAP, NUAK1, PTCH1, and TGFBR2 were validated by qRT-PCR (P < .05) and POSTN, CXCL14, and phosphorylated Smad2/3 were validated by immunohistochemistry (P < .001) as independent predictors of debulking status. The sum of immunohistochemistry intensities for these three proteins provided a tool that classified 92.8% of samples correctly in high- and low-risk groups for suboptimal debulking (area under the curve = 0.89; 95% CI = 0.84 to 0.93).

CONCLUSIONS

Our survival signature provides the most accurate and validated prognostic model for early- and advanced-stage high-grade, serous ovarian cancer. The debulking signature accurately predicts the outcome of cytoreductive surgery, potentially allowing for stratification of patients for primary vs secondary cytoreduction.

Systematic assessment of prognostic gene signatures for breast cancer shows distinct influence of time and ER status

Background

The aim was to assess and compare prognostic power of nine breast cancer gene signatures (Intrinsic, PAM50, 70-gene, 76-gene, Genomic-Grade-Index, 21-gene-Recurrence-Score, EndoPredict, Wound-Response and Hypoxia) in relation to ER status and follow-up time.

Methods

A gene expression dataset from 947 breast tumors was used to evaluate the signatures for prediction of Distant Metastasis Free Survival (DMFS). A total of 912 patients had available DMFS status. The recently published METABRIC cohort was used as an additional validation set.

Results

Survival predictions were fairly concordant across most signatures. Prognostic power declined with follow-up time. During the first 5 years of followup, all signatures except for Hypoxia were predictive for DMFS in ER-positive disease, and 76-gene, Hypoxia and Wound-Response were prognostic in ER-negative disease. After 5 years, the signatures had little prognostic power. Gene signatures provide significant prognostic information beyond tumor size, node status and histological grade.

Conclusions

Generally, these signatures performed better for ER-positive disease, indicating that risk within each ER stratum is driven by distinct underlying biology. Most of the signatures were strong risk predictors for DMFS during the first 5 years of follow-up. Combining gene signatures with histological grade or tumor size, could improve the prognostic power, perhaps also of long-term survival.