BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Efficient Bayesian inference for mechanistic modelling with high-t
 hroughput data - Ruth Baker (University of Oxford)
DTSTART:20230807T101500Z
DTEND:20230807T111500Z
UID:TALK201427@talks.cam.ac.uk
DESCRIPTION:Bayesian methods are routinely used to combine experimental da
 ta with detailed mathematical models to obtain insights into physical phen
 omena. However\, the computational cost of Bayesian computation with detai
 led models has been a notorious problem. While high-throughput data presen
 ts opportunities to calibrate sophisticated models\, comparing large amoun
 ts of data with model simulations quickly becomes computationally prohibit
 ive. Inspired by the method of Stochastic Gradient Descent\, we propose a 
 minibatch approach to approximate Bayesian computation. Through a case stu
 dy of a high-throughput scratch assay experiment\, we show that reliable i
 nference can be performed at a fraction of the computational cost of a tra
 ditional approximate Bayesian inference scheme. By applying a detailed mat
 hematical model of cell movement\, proliferation\, and death to a wide ran
 ge of gene knockdowns\, we characterise the relative contributions of loca
 l cell density-dependent and -independent mechanisms of cell movement and 
 proliferation. Within a screen of 118 gene knockdowns\, we characterise fu
 nctional subgroups of gene knockdowns\, each displaying its own typical co
 mbination of local cell density-dependent and independent movement and pro
 liferation patterns. By comparing these patterns to experimental measureme
 nts of cell counts and wound closure\, we find that density-dependent inte
 ractions play a crucial role in the outcome of the scratch assay.
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR