Deep Learning Imaging: Applications in Genomics and Embryology

A Department of Computational Biology Seminar

Featuring Dr. Iman Hajirasouliha, Associate Professor, Department of Physiology and Biophysics, Weill Cornell Medicine

In this talk, we discuss two deep-learning imaging models with applications in (a) human genome structural variant detection and (b) assessing embryo quality and the prediction of aneuploid embryos.

Structural variants (SV) are a major driver of genetic diversity and disease in the human genome and their discovery is imperative to advances in precision medicine and our understanding of human genetics. Existing SV callers rely on hand-engineered features and heuristics to model SVs, which cannot easily scale to the vast diversity of SV types nor fully harness all the information available in sequencing datasets. Since deep neural networks can learn complex abstractions directly from the data, they offer a promising approach for general SV discovery. Here we present an extensible deep learning framework, Cue, to call and genotype SVs

One challenge in the field of in vitro fertilization is the selection of the most viable embryos for transfer. Manual Morphological quality assessment and morphokinetic analysis both suffer from intra- and inter-observer variability. A third method, pre-implantation genetic testing for aneuploid (PGT-A), has limitations including its invasiveness and cost. We hypothesize that differences in aneuploid and euploid embryos that allow for model-based classification are reflected in morphology, morphokinetics, and associated clinical information. We present a non-invasive and automated method of embryo evaluation that uses deep learning to assess embryo quality and predict embryo ploidy status.