Interpretable machine learning approaches for understanding functional genomics in the human brain

Date of Event

Guest Speaker: Daifeng Wang, PhD. Assistant Professor, Department of Biomedical Informatics

When: Wednesday, November 7, 2018

Time: 3:00 – 4:00PM


Robust phenotype-genotype associations have been established for a number of brain disorders including psychiatric and neurodegenerative diseases. Recent large scientific consortia have generated comprehensive and harmonized functional genomic datasets for the human brain. Integrating and mining these large-scale datasets is both a central priority and a great challenge for understanding the molecular causes of brain disorders and necessitates the development of specialized computational approaches. In this talk, I will present a set of interpretable machine learning approaches that we have recently developed to address this challenge for deciphering functional genomic elements and linkages in the brain and psychiatric disorders. In particular, we have built a comprehensive resource for the human brain using the data of PsychENCODE and other related consortia, including ~5,500 genotype, transcriptome, and chromatin datasets from 1,866 individuals and ~32,000 single-cell datasets. Through the deconvolution using single-cell data, we found that differences in the proportions of cell types explain >85% of the cross-population variation observed. Also, we identified ~79,000 brain-active enhancers and linked them to genes and transcription factors in an extended gene regulatory network. We further identified various QTLs associated with expression, chromatin, splicing and cell-type-proportion changes (e.g., ~2.5M eQTLs comprising ~238K linkage-disequilibrium-independent SNPs). Leveraging our QTLs, Hi-C datasets we connected genes and epigenetic changes to GWAS variants for disorders (e.g., 321 new disease genes for schizophrenia). We also developed a deep-learning model embedding regulatory elements and networks to predict phenotype from genotype. Our model improves disease prediction, highlights key genes and pathways for disorders, and allows imputation of missing transcriptome information from genotype data alone.

Learning Objectives:
1) Learn how to use computational approaches, such as machine learning to identify functional genomic elements interactions for the human brain disorders.
2) I will introduce how to use deep neural network embedding functional genomics to improve the brain disease prediction.
3) I will show potential molecular mechanisms revealed by the deep neural network models for brain disorders and human aging.

Daifeng Wang is an Assistant Professor in Biomedical Informatics at Stony Brook University. He received his Ph.D. in Electrical and Computer Engineering from the University of Texas at Austin in 2011. He joined Gerstein Lab as postdoctoral associate (2012-2015) and associate research scientist (2015-2016) in Yale University. His research has focused on computational biology, bioinformatics, genome informatics. In particular, his lab develops biologically interpretable machine learning approaches for understanding functional genomics in the human brain diseases and cancers.

**CME Credit Available**

The School Of Medicine, State University of New York at Stony Brook, is accredited by the Accreditation Council for Continuing Medical Education to provide continuing medical education for physicians.

The School of Medicine, State University of New York at Stony Brook designates this live activity for a maximum of 1 AMA PRA Category 1 Credit(s) ™.   Physicians should only claim the credit commensurate with the extent of their participation in the activity. Disclosure Policy:  All those in control of CME content are expected to disclose any relevant financial relationship with the provider of commercial products or services discussed in the educational presentation or that have directly supported the CME activity through an educational grant to the sponsoring organization(s).   All commercial relationships that create a conflict with the planners, speakers, author’s control of content must be resolved before the educational activity occurs.

CME Reminder:
·         Everyone MUST register with the CME office by using this link: ALL attendees are tracked by the CME office.
·         All CME credits are now being entered onto the online Stony Brook system.
·         You will be able to view and print your credits yourself whenever you need them.