BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//ox.ac.uk//NONSGML oxford.event//EN
X-WR-TIMEZONE:Europe/London
BEGIN:VTIMEZONE
TZID:Europe/London
X-LIC-LOCATION:Europe/London
BEGIN:DAYLIGHT
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:BST
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:GMT
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
SUMMARY:PEtab SciML: The missing layer for scalable and flexible scientifi
 c machine learning modeling in biology
DTSTART;TZID=Europe/London:20260605T110000
DTEND;TZID=Europe/London:20260605T120000
DTSTAMP:20260604T000910Z
UID:7635735c-8a44-f111-bec7-7c1e52046848
CREATED:20260430T114731Z
DESCRIPTION:Mechanistic ordinary differential equation (ODE) models are a 
 powerful tool to study dynamic biological systems. However\, their predict
 ive power is constrained by gaps\, biases\, and inconsistencies in the lit
 erature. They typically also require quantitative time-lapse data for trai
 ning\, which is time-consuming to collect. At the same time\, machine-lear
 ning approaches can capture complex patterns from data\, but they are ofte
 n harder to interpret and typically require large training datasets. Hybri
 d scientific machine learning (SciML) models offer a promising way to comb
 ine the strengths of both approaches by integrating mechanistic models wit
 h flexible data-driven modules. \nDespite this promise\, the use of SciML 
 in biology remains limited by insufficient infrastructure. Dedicated softw
 are is needed because coding end-to-end differentiable workflows for gradi
 ent-based training of hybrid models is technically challenging. In additio
 n\, model exchange is hindered by the lack of a standardized\, reproducibl
 e format for specifying SciML training problems\, analogous to the PEtab s
 tandard for ODE models. To address these challenges\, we developed PEtab-S
 ciML\, an extension of the PEtab format\, and implemented support for it i
 n the state-of-the-art modeling toolboxes PEtab.jl and AMICI. In this semi
 nar\, I will introduce the PEtab-SciML format. Using real-data examples\, 
 I will show how PEtab-SciML enables the integration of diverse data modali
 ties into dynamic model training\; such as learning the kinetic parameters
  of an ODE model from omics and protein sequence data. I will also show ho
 w it supports machine-learning-based black-boxing of complex model compone
 nts\, such as quarantine strength in an SIR model. Finally\, I will show h
 ow PEtab-SciML enables the use of efficient training strategies\, such as 
 curriculum learning\, that make SciML models easier to train and apply in 
 practice.
LAST-MODIFIED:20260430T114958Z
LOCATION:Mathematical Institute - L4\, L4 Mathematical Institute Woodstock
  Road Oxford Oxfordshire OX2 6GG United Kingdom
SPEAKER:Dr Sebastian Persson (The Francis Crick Institute)
END:VEVENT
END:VCALENDAR