Survey
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project
Download What is a pathway?
Document related concepts
no text concepts found
Transcript
Bayesian Factor Modelling for Omics Data Analysis to Drug Treatments Naruemon Pratanwanich (Ploy) and Pietro Lio’ 28 Oct 2014 Bayesian Factor Modelling for Omics Data Analysis to Drug Treatments Naruemon Pratanwanich (Ploy) and Pietro Lio’ 28 Oct 2014 Outline • Introduction: – Why pathway-based analysis ? – Current tool • Our approach: Bayesian factor modelling • Results: comparison with state-of-the-art methods • Applications: – Drug repositioning – Disease comorbidity – Tissue comparative study Naruemon Pratanwanich (Ploy) Email: [email protected] INTRODUCTION Why pathway-based analysis ? Current tool Naruemon Pratanwanich (Ploy) Email: [email protected] What is a pathway? Naruemon Pratanwanich (Ploy) Email: [email protected] Why pathway-based analysis ? List of DE genes • BRCA1 • MIR151A • SMIM2 • ASAP1-IT1 • FGF14 Naruemon Pratanwanich (Ploy) Pathway profile • GnRH signalling pathway • Caffeine metabolism • MAPK signaling pathway Email: [email protected] Why pathway-based analysis ? List of DE genes • BRCA1 • MIR151A • SMIM2 • ASAP1-IT1 • FGF14 Pathway profile • GnRH signalling pathway • Caffeine metabolism • MAPK signaling pathway Each pathway is individually interpretable Dimension reduction Clarification for one gene – multiple pathways Naruemon Pratanwanich (Ploy) Email: [email protected] BRCA1 functions in multiple pathways Fanconi anemia pathway Naruemon Pratanwanich (Ploy) PI3K-Akt signaling pathway Email: [email protected] Why pathway-based analysis ? List of DE genes • BRCA1 • MIR151A • SMIM2 • ASAP1-IT1 • FGF14 Pathway profile • GnRH signalling pathway • Caffeine metabolism • MAPK signaling pathway Each pathway is individually interpretable Dimension reduction Clarification for one gene – multiple pathways Modest change detection Naruemon Pratanwanich (Ploy) Email: [email protected] BRCA1 functions in multiple pathways BRCA1 BRCA1 Fanconi anemia pathway Naruemon Pratanwanich (Ploy) PI3K-Akt signaling pathway Email: [email protected] Current tool Gene set enrichment analysis (GSEA) Do not take pathway dependency into account GSEA p-value = ? Broad Institute 2005 One sample and one pathway at a time ! Naruemon Pratanwanich (Ploy) Email: [email protected] OUR APPROACH Bayesian factor modelling Naruemon Pratanwanich (Ploy) Email: [email protected] Drug treatments Disease perturbations A B Naruemon Pratanwanich (Ploy) Email: [email protected] Gaussian Markov Random Fields Naruemon Pratanwanich (Ploy) Email: [email protected] Assumptions • Perturbation → Certain pathways → Gene expression levels • Association strengths between genes and pathways are different • Pathways are NOT independent Naruemon Pratanwanich (Ploy) Email: [email protected] Naruemon Pratanwanich (Ploy) Email: [email protected] Naruemon Pratanwanich (Ploy) Email: [email protected] Naruemon Pratanwanich (Ploy) Email: [email protected] Naruemon Pratanwanich (Ploy) Email: [email protected] RESULTS Comparison with state-of-the-art methods using a set of known drug-pathway pairs from an external database Naruemon Pratanwanich (Ploy) Email: [email protected] drug i drug j drug i drug j Most probable responsive pathways Pathway rank profiles Naruemon Pratanwanich (Ploy) Email: [email protected] drug i drug j drug i drug j Most probable responsive pathways Pathway rank profiles Naruemon Pratanwanich (Ploy) Email: [email protected] Bad: Drugs with a few pathways Good: Drugs with many pathways Naruemon Pratanwanich (Ploy) Email: [email protected] APPLICATIONS OF PATHWAY NETWORK Disease comorbidity Drug repositioning Tissue comparative study Naruemon Pratanwanich (Ploy) Email: [email protected] Disease comorbidity Hepatitis B Thyroid cancer FCεRI NF-κB Naruemon Pratanwanich (Ploy) Email: [email protected] Disease comorbidity Hepatitis B Thyroid cancer FCεRI NF-κB Naruemon Pratanwanich (Ploy) Email: [email protected] Drug repositioning Colorectal cancer Naruemon Pratanwanich (Ploy) GnRH Email: [email protected] Drug repositioning Colorectal cancer Naruemon Pratanwanich (Ploy) GnRH Email: [email protected] Drug repositioning Colorectal cancer Naruemon Pratanwanich (Ploy) GnRH Email: [email protected] Drug repositioning Colorectal cancer Naruemon Pratanwanich (Ploy) GnRH Email: [email protected] Tissue comparative analysis MCF7 – Breast cancer PC3 – Prostate cancer DE genes Compare FcƐRI NF-κB FcƐRI Crosstalks NF-κB Compare Rank 10th Naruemon Pratanwanich (Ploy) Rank 11,765th Email: [email protected] Conclusions • Identify pathway responsiveness and pathway crosstalks • Generate informative hypothesis related to molecular processes based on pathways • Future issues: sign flipping, interaction within pathways Naruemon Pratanwanich (Ploy) Email: [email protected] Confirm that pathways are not independent ! FacPad Our method Naruemon Pratanwanich (Ploy) GSEA Email: [email protected] Naruemon Pratanwanich (Ploy) Email: [email protected] Naruemon Pratanwanich (Ploy) Email: [email protected]
Related documents