skip to main content


Title: mClass: Cancer Type Classification with Somatic Point Mutation Data
Cancer is a complex disease associated with abnormal DNA mutations. Not all tumors are cancerous and not all cancers are the same. Correct cancer type diagnosis can indicate the most effective drug therapy and increase survival rate. At the molecular level, it has been shown that cancer type classification can be carried out from the analysis of somatic point mutation. However, the high dimensionality and sparsity of genomic mutation data, coupled with its small sample size has been a hindrance in accurate classification of cancer. We address these problems by introducing a novel classification method called mClass that accounts for the sparsity of the data. mClass is a feature selection method that ranks genes based on their similarity across samples and employs their normalized mutual information to determine the set of genes that provide optimal classification accuracy. Experimental results on TCGA datasets show that mClass significantly improves testing accuracy compared to DeepGene, which is the state-of-the-art in cancer-type classification based on somatic mutation data. In addition, when compared with other cancer gene prediction tools, the set of genes selected by mClass contains the highest number of genes in top 100 genes listed in the Cancer Gene Census. mClass is available at https://github.com/mdahasan/mClass.  more » « less
Award ID(s):
1526742 1543963
NSF-PAR ID:
10094705
Author(s) / Creator(s):
Date Published:
Journal Name:
RECOMB-CG 2018 - International Conference on Comparative Genomics
Page Range / eLocation ID:
131-145
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract

    Identification of a few cancer driver mutation genes from a much larger number of passenger mutation genes in cancer samples remains a highly challenging task. Here, a novel method for distinguishing the driver genes from the passenger genes by effective integration of somatic mutation data and molecular interaction data using a maximal mutational impact function (MaxMIF) is presented. When evaluated on six somatic mutation datasets of Pan‐Cancer and 19 datasets of different cancer types from TCGA, MaxMIF almost always significantly outperforms all the existing state‐of‐the‐art methods in terms of predictive accuracy, sensitivity, and specificity. It recovers about 30% more known cancer genes in 500 top‐ranked candidate genes than the best among the other tools evaluated. MaxMIF is also highly robust to data perturbation. Intriguingly, MaxMIF is able to identify potential cancer driver genes, with strong experimental data support. Therefore, MaxMIF can be very useful for identifying or prioritizing cancer driver genes in the increasing number of available cancer genomic data.

     
    more » « less
  2. Abstract Motivation Somatic mutations result from processes related to DNA replication or environmental/lifestyle exposures. Knowing the activity of mutational processes in a tumor can inform personalized therapies, early detection, and understanding of tumorigenesis. Computational methods have revealed 30 validated signatures of mutational processes active in human cancers, where each signature is a pattern of single base substitutions. However, half of these signatures have no known etiology, and some similar signatures have distinct etiologies, making patterns of mutation signature activity hard to interpret. Existing mutation signature detection methods do not consider tumor-level clinical/demographic (e.g. smoking history) or molecular features (e.g. inactivations to DNA damage repair genes). Results To begin to address these challenges, we present the Tumor Covariate Signature Model (TCSM), the first method to directly model the effect of observed tumor-level covariates on mutation signatures. To this end, our model uses methods from Bayesian topic modeling to change the prior distribution on signature exposure conditioned on a tumor’s observed covariates. We also introduce methods for imputing covariates in held-out data and for evaluating the statistical significance of signature-covariate associations. On simulated and real data, we find that TCSM outperforms both non-negative matrix factorization and topic modeling-based approaches, particularly in recovering the ground truth exposure to similar signatures. We then use TCSM to discover five mutation signatures in breast cancer and predict homologous recombination repair deficiency in held-out tumors. We also discover four signatures in a combined melanoma and lung cancer cohort—using cancer type as a covariate—and provide statistical evidence to support earlier claims that three lung cancers from The Cancer Genome Atlas are misdiagnosed metastatic melanomas. Availability and implementation TCSM is implemented in Python 3 and available at https://github.com/lrgr/tcsm, along with a data workflow for reproducing the experiments in the paper. Supplementary information Supplementary data are available at Bioinformatics online. 
    more » « less
  3. Background

    Gene co‐expression and differential co‐expression analysis has been increasingly used to study co‐functional and co‐regulatory biological mechanisms from large scale transcriptomics data sets.

    Methods

    In this study, we develop a nonparametric approach to identify hub genes and modules in a large co‐expression network with low computational and memory cost, namely MRHCA.

    Results

    We have applied the method to simulated transcriptomics data sets and demonstrated MRHCA can accurately identify hub genes and estimate size of co‐expression modules. With applying MRHCA and differential co‐expression analysis toE. coliand TCGA cancer data, we have identified significant condition specific activated genes inE. coliand distinct gene expression regulatory mechanisms between the cancer types with high copy number variation and small somatic mutations.

    Conclusion

    Our analysis has demonstrated MRHCA can (i) deal with large association networks, (ii) rigorously assess statistical significance for hubs and module sizes, (iii) identify co‐expression modules with low associations, (iv) detect small and significant modules, and (v) allow genes to be present in more than one modules, compared with existing methods.

     
    more » « less
  4. Abstract

    Identification and characterization of somatic mutations in cancer have important prognostication and treatment implications. Genes encoding the Nuclear factor (erythroid-derived 2)-like 2 (NRF2) transcription factor and its negative regulator, Kelch-like ECH-associated protein 1 (KEAP1), are frequently mutated in cancer. These mutations drive constitutive NRF2 activation and correlate with poor prognosis. Despite its apparent significance, a comprehensive catalogue of somatic NRF2 mutations across different tumor types is still lacking. Here, we catalogue NRF2 mutations in The Cancer Genome Atlas (TCGA) database. 226 unique NRF2-mutant tumors were identified from 10,364 cases. NRF2 mutations were found in 21 out of the 33 tumor types. A total of 11 hotspots were identified. Of these, mutation to the R34 position was most frequent. Notably, R34 and D29 mutations were overrepresented in bladder, lung, and uterine cancers. Analyses of corresponding RNA sequencing data using a de novo derived gene expression classifier showed that the R34 mutations drive constitutive NRF2 activation with a selection pressure biased against the formation of R34L. Of all R34 mutants, R34L conferred the least degree of protein stabilization, suggesting a pro-tumor NRF2 half-life threshold. Our findings offer a comprehensive catalogue of NRF2 mutations in cancer that can help prognostication and NRF2 research.

     
    more » « less
  5. Finding genes biologically directly or indirectly related to lung cancer has been drawing much attention, and many genes directly related to lung cancer have been reported. However, it has not been confirmed whether those published 'key' genes are truly critical to lung cancer formation, i.e., they may be with very limited useful information. As a result, finding essential genes remains a challenging lung cancer research problem. Using a recently developed competing linear factor analysis method in differentially expressed gene detection, we advance the study of lung cancer critical genes detection to a uniformly informative level. A set of common four genes and their functional effects are detected to be differentially expressed in tumor and non- tumor samples with 100% sensitivity and 100% specificity in one study of lung adenocarcinoma (LUAD) and one study of squamous cell lung cancers (LUSC) (two North American cohorts with 20429 genes, 576 and 552 samples respectively). Two additional analyses also gain accuracy of 97.8% sensitivity and 100% specificity in one study of non-small cell lung carcinomas (NSCLC, a European cohort with 20356 genes and 156 samples), and an accuracy of 100% sensitivity and 95% specificity (1 out of 20 non-tumor samples) in one study of ALK-positive and EGFR/KRAS/ALK-negative lung adenocarcinomas (LUAD, a Japanese cohort with 20356 genes and 224 samples). There are some common genes, but different functional effects, within each set of four genes among two North American cohorts and a European cohort and among North American cohorts and the Japanese cohort. These results show the four-gene-based classifiers are robust with different types of lung cancers and different race cohorts and accurate. The functional effects of four genes disclose significantly other mechanisms (mysteries) between LUAD and LUSC. These sets of four genes and their functional effects are considered to be essential for lung cancer studies and practice. These genes' functional effects naturally classify patients into different groups (more than seven subtypes). Subtype information is useful for personalized therapies. The new findings can motivate new lung cancer research in more focused and targeted directions to save lives, protect people, and reduce enormous economic costs in research and lung cancer treatments. 
    more » « less