Cox-nnet: an artificial neural network Cox regression for prognosis prediction

Artificial neural networks (ANN) are computing architectures with massively parallel interconnections of simple neurons and has been applied to biomedical fields such as imaging analysis and diagnosis. We have developed a new ANN framework called Cox-nnet to predict patient prognosis from high throughput transcriptomics data. In over 10 TCGA RNA-Seq data sets, Cox-nnet achieves a statistically significant increase in predictive accuracy, compared to the other three methods including Cox-proportional hazards (Cox-PH), Random Forests Survival and CoxBoost. Cox-nnet also offers richer biological information, from both pathway and gene levels. The outputs from the hidden layer node can be utilized as a new approach for survival-sensitive dimension reduction. In summary, we have developed a new method for more accurate and efficient prognosis prediction, with functional biological insights. The source code is freely available at https://github.com/lanagarmire/cox-nnet.

[1]  M. Gevrey,et al.  Review and comparison of methods to study the contribution of variables in artificial neural network models , 2003 .

[2]  Joshua Zhexue Huang,et al.  Unbiased Feature Selection in Learning Random Forests for High-Dimensional Data , 2015, TheScientificWorldJournal.

[3]  Susumu Goto,et al.  KEGG: Kyoto Encyclopedia of Genes and Genomes , 2000, Nucleic Acids Res..

[4]  Harvey Risch,et al.  Meta-dimensional data integration identifies critical pathways for susceptibility, tumorigenesis and progression of endometrial cancer , 2016, Oncotarget.

[5]  H. Tarhan,et al.  p53 Mutations and Other Prognostic Factors of Renal Cell Carcinoma , 2001, Urologia Internationalis.

[6]  Russell G. Death,et al.  An accurate comparison of methods for quantifying variable importance in artificial neural networks using simulated data , 2004 .

[7]  Tom C. Freeman,et al.  Improved grading and survival prediction of human astrocytic brain tumors by artificial neural network analysis of gene expression microarray data , 2008, Molecular Cancer Therapeutics.

[8]  A. Adjei,et al.  Advances in Targeting the Ras/Raf/MEK/Erk Mitogen-Activated Protein Kinase Cascade with MEK Inhibitors for Cancer Therapy , 2008, Clinical Cancer Research.

[9]  D Faraggi,et al.  A neural network model for survival data. , 1995, Statistics in medicine.

[10]  Hemant Ishwaran,et al.  Random Survival Forests , 2008, Wiley StatsRef: Statistics Reference Online.

[11]  R. Medeiros,et al.  Influence of survivin (BIRC5) and caspase-9 (CASP9) functional polymorphisms in renal cell carcinoma development: a study in a southern European population , 2013, Molecular Biology Reports.

[12]  J. Koziol,et al.  The Concordance Index C and the Mann–Whitney Parameter Pr(X>Y) with Randomly Censored Data , 2009, Biometrical journal. Biometrische Zeitschrift.

[13]  Lana X. Garmire,et al.  A Novel Model to Combine Clinical and Pathway-Based Transcriptomic Information for the Prognosis Prediction of Breast Cancer , 2014, PLoS Comput. Biol..

[14]  Erwin G. Van Meir,et al.  Emerging roles for the BAI1 protein family in the regulation of phagocytosis, synaptogenesis, neurovasculature, and tumor development , 2011, Journal of Molecular Medicine.

[15]  Martin J. Wainwright,et al.  Noisy matrix decomposition via convex relaxation: Optimal rates in high dimensions , 2011, ICML.

[16]  Daniel B. Mark,et al.  TUTORIAL IN BIOSTATISTICS MULTIVARIABLE PROGNOSTIC MODELS: ISSUES IN DEVELOPING MODELS, EVALUATING ASSUMPTIONS AND ADEQUACY, AND MEASURING AND REDUCING ERRORS , 1996 .

[17]  Yusuke Nakamura,et al.  A novel brain-specific p53-target gene, BAI1, containing thrombospondin type 1 repeats inhibits experimental angiogenesis , 1997, Oncogene.

[18]  Keith T Flaherty,et al.  Phase II and biomarker study of the dual MET/VEGFR2 inhibitor foretinib in patients with papillary renal cell carcinoma. , 2013, Journal of clinical oncology : official journal of the American Society of Clinical Oncology.

[19]  T. Fujioka,et al.  Brain-specific angiogenesis inhibitor 1 is a putative factor for inhibition of neovascular formation in renal cell carcinoma. , 2011, The Journal of urology.

[20]  L. V. van't Veer,et al.  Cross‐validated Cox regression on microarray gene expression data , 2006, Statistics in medicine.