BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Deep Bayesian Recurrent Neural Networks for Somatic Variant Callin
 g - Omar Darwish - DAMTP and Geoffroy Dubourg-Felonneau - CCG
DTSTART:20200305T130000Z
DTEND:20200305T143000Z
UID:TALK136078@talks.cam.ac.uk
CONTACT:James Fergusson
DESCRIPTION:The genomic profile underlying an individual tumor can be high
 ly informative in the creation of a personalized cancer treatment strategy
  for a given patient\; a practice known as precision oncology. This involv
 es next generation sequencing (NGS) of a tumor sample and the subsequent i
 dentification of genomic abberations\, such as  somatic mutations\, to pro
 vide potential candidates of targeted therapy. The identification of these
  abberations from sequencing artifacts can be seen as a classification tas
 k. This has been previously broached with many different supervised machin
 e learning methods\, including neural networks. However\, these neural net
 works have thus far not been tailored to give any indication of confidence
  in the mutation call\, meaning an oncologist could be targeting a mutatio
 n with a low probability of being real. To address this\, we present a dee
 p bayesian recurrent neural network for cancer variant calling\, which sho
 ws no degradation in performance compared to standard neural networks but 
 yet returnss a measure of the confidence that reflects its performance on 
 out-of-distribution data. We hope this approach can be incorporated into s
 oftware used by oncologists to provide statistical confidence in precision
  oncology treatment choices.
LOCATION:Kavli Large Meeting Room\, Kavli Building
END:VEVENT
END:VCALENDAR