BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20260603T152131EDT-2573EuLuvL@132.216.98.100
DTSTAMP:20260603T192131Z
DESCRIPTION:Abstract\n\nComputationally designing personalized treatment pl
 ans to increase a cancer patient's chances of recovery using their molecul
 ar profiles has been one of the major objectives of precision cancer medic
 ine. Despite the advancement of high-throughput sequencing and artificial 
 intelligence\, drug response prediction has remained a challenging task. T
 his thesis presents novel methodologies for predicting responses to drug t
 reatments\, addressing challenges such as limited clinical data and drug-s
 pecific biases. Leveraging available datasets\, we explored the utility of
  different information modalities into predictive models.\n\nFirst\, we fo
 cused on clinical drug response prediction using only preclinical data. Th
 is stemmed from the current situation of cancer drug response datasets\, w
 herein drug responses for preclinical cancer cell line (CCL) samples treat
 ed with hundreds of drugs are widely available\, while clinical drug respo
 nses of tumors are only available in small patient cohorts for a handful o
 f drugs. We developed a deep learning pipeline that leverages tissue infor
 mation to bridge discrepancies between CCL and tumor samples\, enabling mo
 dels to distinguish between sensitive and resistant patients.\n\nWe then v
 entured towards improving drug representation using knowledge graphs compo
 sed of CCLs\, drugs\, and genes. Unlike previous methods that solely rely 
 on the structural properties of drug molecules\, we integrated additional 
 response-relevant information\, such as molecular profiles of extremely se
 nsitive/resistant CCLs\, CRISPR gene effects\, and drug targets. Our analy
 ses demonstrated superior performance compared to existing methods and bas
 eline approaches.\n\nBeyond drug response prediction\, we also identified 
 potential biomarkers of drug response for each model that we presented. Th
 is not only enhances model interpretability\, but also produces data-drive
 n hypotheses. Many implicated genes and pathways were supported by literat
 ure\, and in some cases\, experimentally validated. We introduced a graph-
 based interpretation method to provide further insights and visualize the 
 prediction process at a high level.\n\nThe content of this thesis not only
  improve drug response prediction but also sheds light on potential therap
 eutic targets\, contributing to the advancement of precision cancer medici
 ne.\n
DTSTART:20240730T180000Z
DTEND:20240730T200000Z
LOCATION:Room 603\, McConnell Engineering Building\, CA\, QC\, Montreal\, H
 3A 0E9\, 3480 rue University
SUMMARY:PhD Defence of David Hostallero - Applications of Deep Learning and
  Graph Representation Learning in Precision Cancer Medicine
URL:https://www.mcgill.ca/ece/channels/event/phd-defence-david-hostallero-a
 pplications-deep-learning-and-graph-representation-learning-precision-3580
 50
END:VEVENT
END:VCALENDAR