Leveraging TCGA gene expression data to build predictive models for cancer drug response.

BACKGROUND: Machine learning has been utilized to predict cancer drug response from multi-omics data generated from sensitivities of cancer cell lines to different therapeutic compounds. Here, we build machine learning models using gene expression data from patients' primary tumor tissues to predict whether a patient will respond positively or negatively to two chemotherapeutics: 5-Fluorouracil and Gemcitabine. RESULTS: We focused on 5-Fluorouracil and Gemcitabine because based on our exclusion criteria, they provide the largest numbers of patients within TCGA. Normalized gene expression data were clustered and used as the input features for the study. We used matching clinical trial data to ascertain the response of these patients via multiple classification methods. Multiple clustering and classification methods were compared for prediction accuracy of drug response. Clara and random forest were found to be the best clustering and classification methods, respectively. The results show our models predict with up to 86% accuracy; despite the study's limitation of sample size. We also found the genes most informative for predicting drug response were enriched in well-known cancer signaling pathways and highlighted their potential significance in chemotherapy prognosis. CONCLUSIONS: Primary tumor gene expression is a good predictor of cancer drug response. Investment in larger datasets containing both patient gene expression and drug response is needed to support future work of machine learning models. Ultimately, such predictive models may aid oncologists with making critical treatment decisions.

Tumor suppressor genes and allele-specific expression: mechanisms and significance.

Recent findings indicate that allele-specific expression (ASE) at specific cancer driver gene loci may be of importance in onset/progression of the disease. Of particular interest are loss-of-function (LOF) of tumor suppressor gene (TSGs) alleles. While LOF tumor suppressor mutations are typically considered to be recessive, if these mutant alleles can be significantly differentially expressed relative to wild-type alleles in heterozygotes, the clinical consequences could be significant. LOF TSG alleles are shown to be segregating at high frequencies in world-wide populations of normal/healthy individuals. Matched sets of normal and tumor tissues isolated from 233 cancer patients representing four diverse tumor types demonstrate functionally important changes in patterns of ASE in individuals heterozygous for LOF TSG alleles associated with cancer onset/progression. While a variety of molecular mechanisms were identified as potentially contributing to changes in ASE patterns in cancer, changes in DNA copy number and allele-specific alternative splicing possibly mediated by antisense RNA emerged as predominant factors. In conclusion, LOF TSGs are segregating in human populations at significant frequencies indicating that many otherwise healthy individuals are at elevated risk of developing cancer. Changes in ASE between normal and cancer tissues indicates that LOF TSG alleles may contribute to cancer onset/progression even when heterozygous with wild-type functional alleles.

An atlas of transposable element-derived alternative splicing in cancer.

Transposable element (TE)-derived sequences comprise more than half of the human genome, and their presence has been documented to alter gene expression in a number of different ways, including the generation of alternatively spliced transcript isoforms. Alternative splicing has been associated with tumorigenesis for a number of different cancers. The objective of this study was to broadly characterize the role of human TEs in generating alternatively spliced transcript isoforms in cancer. To do so, we screened for the presence of TE-derived sequences co-located with alternative splice sites that are differentially used in normal versus cancer tissues. We analysed a comprehensive set of alternative splice variants characterized for 614 matched normal-tumour tissue pairs across 13 cancer types, resulting in the discovery of 4820 TE-generated alternative splice events distributed among 723 cancer-associated genes. Short interspersed nuclear elements (Alu) and long interspersed nuclear elements (L1) were found to contribute the majority of TE-generated alternative splice sites in cancer genes. A number of cancer-associated genes, including MYH11, WHSC1 and CANT1, were shown to have overexpressed TE-derived isoforms across a range of cancer types. TE-derived isoforms were also linked to cancer-specific fusion transcripts, suggesting a novel mechanism for the generation of transcriptome diversity via trans-splicing mediated by dispersed TE repeats. This article is part of a discussion meeting issue 'Crossroads between transposons and gene regulation'.

Machine learning predicts individual cancer patient responses to therapeutic drugs with high accuracy.

Precision or personalized cancer medicine is a clinical approach that strives to customize therapies based upon the genomic profiles of individual patient tumors. Machine learning (ML) is a computational method particularly suited to the establishment of predictive models of drug response based on genomic profiles of targeted cells. We report here on the application of our previously established open-source support vector machine (SVM)-based algorithm to predict the responses of 175 individual cancer patients to a variety of standard-of-care chemotherapeutic drugs from the gene-expression profiles (RNA-seq or microarray) of individual patient tumors. The models were found to predict patient responses with >80% accuracy. The high PPV of our algorithms across multiple drugs suggests a potential clinical utility of our approach, particularly with respect to the identification of promising second-line treatments for patients failing standard-of-care first-line therapies.

Population and clinical genetics of human transposable elements in the (post) genomic era.

Recent technological developments-in genomics, bioinformatics and high-throughput experimental techniques-are providing opportunities to study ongoing human transposable element (TE) activity at an unprecedented level of detail. It is now possible to characterize genome-wide collections of TE insertion sites for multiple human individuals, within and between populations, and for a variety of tissue types. Comparison of TE insertion site profiles between individuals captures the germline activity of TEs and reveals insertion site variants that segregate as polymorphisms among human populations, whereas comparison among tissue types ascertains somatic TE activity that generates cellular heterogeneity. In this review, we provide an overview of these new technologies and explore their implications for population and clinical genetic studies of human TEs. We cover both recent published results on human TE insertion activity as well as the prospects for future TE studies related to human evolution and health.

Patterns of Transposable Element Expression and Insertion in Cancer.

Human transposable element (TE) activity in somatic tissues causes mutations that can contribute to tumorigenesis. Indeed, TE insertion mutations have been implicated in the etiology of a number of different cancer types. Nevertheless, the full extent of somatic TE activity, along with its relationship to tumorigenesis, have yet to be fully explored. Recent developments in bioinformatics software make it possible to analyze TE expression levels and TE insertional activity directly from transcriptome (RNA-seq) and whole genome (DNA-seq) next-generation sequence data. We applied these new sequence analysis techniques to matched normal and primary tumor patient samples from the Cancer Genome Atlas (TCGA) in order to analyze the patterns of TE expression and insertion for three cancer types: breast invasive carcinoma, head and neck squamous cell carcinoma, and lung adenocarcinoma. Our analysis focused on the three most abundant families of active human TEs: Alu, SVA, and L1. We found evidence for high levels of somatic TE activity for these three families in normal and cancer samples across diverse tissue types. Abundant transcripts for all three TE families were detected in both normal and cancer tissues along with an average of ~80 unique TE insertions per individual patient/tissue. We observed an increase in L1 transcript expression and L1 insertional activity in primary tumor samples for all three cancer types. Tumor-specific TE insertions are enriched for private mutations, consistent with a potentially causal role in tumorigenesis. We used genome feature analysis to investigate two specific cases of putative cancer-causing TE mutations in further detail. An Alu insertion in an upstream enhancer of the CBL tumor suppressor gene is associated with down-regulation of the gene in a single breast cancer patient, and an L1 insertion in the first exon of the BAALC gene also disrupts its expression in head and neck squamous cell carcinoma. Our results are consistent with widespread somatic activity of human TEs leading to numerous insertion mutations that can contribute to tumorigenesis in a variety of tissues.