BDKANN – Biological Domain Knowledge-based Artificial Neural Network for drug response prediction

One of the main goals of precision oncology is to predict the response of a patient to a given cancer treatment based on their genomic profile. Although current models for drug response prediction are becoming more accurate, they are also "black boxes" and cannot explain their predictions, which is of particular importance in cancer treatment. Many models also do not leverage prior biological knowledge, such as the hierarchical information on how proteins form complexes and act together in pathways. In this work, we use this prior biological knowledge to form the architecture of a deep neural network to predict cancer drug response from cell line expression data. We find that our approach not only has low prediction error compared to baseline models but also allows for meaningful interpretation of the network. These interpretations can both explain predictions made and discover novel connections in the biological knowledge that could lead to new hypotheses about mechanisms of drug action.

[1]  M. Viola,et al.  Expression of ras oncogene p21 in prostate cancer. , 1986, The New England journal of medicine.

[2]  S. Rodenhuis,et al.  Clinical significance of ras oncogene activation in human lung cancer. , 1992, Cancer research.

[3]  T. Golub,et al.  A method for high-throughput gene expression signature analysis , 2006, Genome Biology.

[4]  Victoria Bolós,et al.  Notch signaling in development and cancer. , 2007, Endocrine reviews.

[5]  B. Taylor,et al.  Transcriptional pathway signatures predict MEK addiction and response to selumetinib (AZD6244). , 2010, Cancer research.

[6]  A. Hill,et al.  Breast cancer cell migration is regulated through junctional adhesion molecule-A-mediated activation of Rap1 GTPase , 2011, Breast Cancer Research.

[7]  S. Ramaswamy,et al.  Systematic identification of genomic markers of drug sensitivity in cancer cells , 2012, Nature.

[8]  Adam A. Margolin,et al.  The Cancer Cell Line Encyclopedia enables predictive modeling of anticancer drug sensitivity , 2012, Nature.

[9]  N. Cox,et al.  Clinical drug response can be predicted using baseline gene expression levels and in vitro drug sensitivity in cell lines , 2014, Genome Biology.

[10]  V. Tuohy.  Retired self-proteins as vaccine targets for primary immunoprevention of adult-onset cancers , 2014, Expert review of vaccines.

[11]  Alexander Binder,et al.  On Pixel-Wise Explanations for Non-Linear Classifier Decisions by Layer-Wise Relevance Propagation , 2015, PloS one.

[12]  Joshua M. Korn,et al.  High-throughput screening using patient-derived tumor xenografts to predict clinical trial drug response , 2015, Nature Medicine.

[13]  Zhenzhen Huang,et al.  Identification of Gene Expression Pattern Related to Breast Cancer Survival Using Integrated TCGA Datasets and Genomic Tools , 2015, BioMed research international.

[14]  Michael P. Morrissey,et al.  Pharmacogenomic agreement between two cancer cell line data sets , 2015, Nature.

[15]  P. Neven,et al.  Fulvestrant with or without selumetinib, a MEK 1/2 inhibitor, in breast cancer progressing after aromatase inhibitor therapy: a multicentre randomised placebo-controlled double-blind phase II trial, SAKK 21/08. , 2015, European journal of cancer.

[16]  Joshua A. Bittker,et al.  Correlating chemical sensitivity and basal gene expression reveals mechanism of action , 2015, Nature chemical biology.

[17]  Andrew H. Beck,et al.  Revisiting inconsistency in large pharmacogenomic studies , 2015, bioRxiv.

[18]  David T. W. Jones,et al.  EPN-30YAP1-MAMLD1 FUSIONS ALONE ARE SUFFICIENT TO FORM SUPRATENTORIAL EPENDYMOMA-LIKE TUMORS IN MICE , 2016 .

[19]  Emanuel J. V. Gonçalves,et al.  A Landscape of Pharmacogenomic Interactions in Cancer , 2016, Cell.

[20]  Andrew H. Beck,et al.  PharmacoGx: an R package for analysis of large pharmacogenomic datasets , 2015, Bioinform..

[21]  F. Sotgia,et al.  Mitochondrial markers predict recurrence, metastasis and Tamoxifen-resistance in breast cancer patients: Early detection of treatment failure with companion diagnostics. , 2017, Oncotarget.

[22]  Avanti Shrikumar,et al.  Learning Important Features Through Propagating Activation Differences , 2017, ICML.

[23]  Lovelace J. Luquette,et al.  A Pan-Cancer Proteogenomic Atlas of PI3K/AKT/mTOR Pathway Alterations. , 2017, Cancer cell.

[24]  Scott Lundberg,et al.  A Unified Approach to Interpreting Model Predictions , 2017, NIPS.

[25]  Angela N. Brooks,et al.  A Next Generation Connectivity Map: L1000 Platform and the First 1,000,000 Profiles , 2017, Cell.

[26]  S. Al-Bahlani,et al.  Calpain-1 Expression in Triple-Negative Breast Cancer: A Potential Prognostic Factor Independent of the Proliferative/Apoptotic Index , 2017, BioMed research international.

[27]  Tianyu Kang,et al.  A biological network-based regularized artificial neural network model for robust phenotype prediction from gene expression data , 2017, BMC Bioinformatics.

[28]  Michael Q. Ding,et al.  Precision Oncology beyond Targeted Therapy: Combining Omics Data with Machine Learning Matches the Majority of Cancer Cells to Effective Therapeutics , 2017, Molecular Cancer Research.

[29]  Yufei Huang,et al.  Predicting drug response of tumors from integrated genomic profiles by deep neural networks , 2018, BMC Medical Genomics.

[30]  Karsten M. Borgwardt,et al.  Kernelized rank learning for personalized drug recommendation , 2017, Bioinform..

[31]  Roded Sharan,et al.  Using deep learning to model the hierarchical structure and function of a cell , 2018, Nature Methods.

[32]  Thawfeek M. Varusai,et al.  The Reactome Pathway Knowledgebase , 2017, Nucleic acids research.

[33]  Tero Aittokallio,et al.  Machine learning and feature selection for drug response prediction in precision oncology applications , 2018, Biophysical Reviews.

[34]  V. Prasad,et al.  Estimation of the Percentage of US Patients With Cancer Who Benefit From Genome-Driven Oncology , 2018, JAMA oncology.

[35]  Zhaleh Safikhani,et al.  PharmacoDB: an integrative database for mining in vitro anticancer drug screening studies , 2017, bioRxiv.

[36]  Jie Hao,et al.  PASNet: pathway-associated sparse deep neural network for prognosis prediction from high-throughput data , 2018, BMC Bioinformatics.

[37]  Marcel J. T. Reinders,et al.  PRECISE: a domain adaptation approach to transfer predictors of drug response from pre-clinical models to tumors , 2019, bioRxiv.

[38]  Adel Javanmard,et al.  Dynamic Pricing in High-Dimensions , 2016, J. Mach. Learn. Res..

[39]  C. Collins,et al.  MOLI: multi-omics late integration with deep neural networks for drug response prediction , 2019, bioRxiv.

[40]  Junhui Qi,et al.  Attenuation of MAMLD1 Expression Suppresses the Growth and Migratory Properties of Gonadotroph Pituitary Adenomas , 2019, Pathology & Oncology Research.

[41]  Martin Ester,et al.  MOLI: multi-omics late integration with deep neural networks for drug response prediction , 2019, Bioinform..

[42]  Klaus-Robert Müller,et al.  iNNvestigate neural networks! , 2018, J. Mach. Learn. Res..

[43]  A. Valencia,et al.  Unveiling new disease, pathway, and gene associations via multi-scale neural network , 2019, PloS one.

[44]  Martin Ester,et al.  AITL: Adversarial Inductive Transfer Learning with input and output space adaptation for pharmacogenomics , 2020, bioRxiv.