Abstract

Using Deep Learning to Predict Activity and Drug Response of Unknown Mutations

By applying deep learning to images of mutated cells interacting with drugs, we can predict how cells with unknown mutations respond to these drugs. State-of-the-art cancer treatment starts with sequencing a tumor biopsy. Alas, we don't know how most mutations act or respond to drugs. Therefore, the treating physician cannot integrate this data into the treating protocol. To solve this, ...


Session No: SIL8108
Speaker: Michael Vidne
Type:

AI in Healthcare

Date: Thursday - October 18, 2018 02:00 PM - 02:45 PM
Location: Hall M
Topics: Deep Learning and AI, AI in Healthcare
Industry: Healthcare & Life Sciences

By applying deep learning to images of mutated cells interacting with drugs, we can predict how cells with unknown mutations respond to these drugs. State-of-the-art cancer treatment starts with sequencing a tumor biopsy. Alas, we don't know how most mutations act or respond to drugs. Therefore, the treating physician cannot integrate this data into the treating protocol. To solve this, we synthesize patients' mutated genes from the sequencing data and transfect the mutated genes into live-cells. The cells express the mutated genes, and we scan the images under the microscope. We use our dataset of images of over 8 billion cells expressing specific mutations to train a deep learning network to classify a mutation and predict its level of activity. The model classifies the images and predicts the mutations' response to different drugs.