BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:Europe/Stockholm
X-LIC-LOCATION:Europe/Stockholm
BEGIN:DAYLIGHT
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
TZNAME:CEST
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
TZNAME:CET
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20250822T115804Z
LOCATION:Room 5.2D02
DTSTART;TZID=Europe/Stockholm:20250617T120000
DTEND;TZID=Europe/Stockholm:20250617T123000
UID:submissions.pasc-conference.org_PASC25_sess134_msa173@linklings.com
SUMMARY:MLDocking: Accelerated Drug Discovery with Transformer-Based Surro
 gate Models and In-Memory Workflows on Heterogeneous HPC Systems
DESCRIPTION:Riccardo Balin, Christine Simpson, Harikrishna Tummalapalli, A
 rchit Vasan, and Venkat Vishwanath (Argonne National Laboratory) and Kent 
 Lee, Yian Chen, Nick Hill, Colin Wahl, and Pete Mendygral (HPE)\n\nThe use
  of AI in drug discovery workflows has accelerated the task of screening b
 illions of molecules to identify top candidates for binding to particular 
 proteins. Typically, these workflows are composed of distinct tasks run se
 quentially on HPC clusters to iteratively screen through the list of compo
 unds, identify top candidates, perform molecular dynamics simulations, and
  fine-tune the AI surrogate. The sequential nature of these offline workfl
 ows results in multiple job submissions with long queue times and heavy us
 e of the parallel file system. In this talk, we present MLDocking – an aut
 omated drug discovery workflow which leverages a novel distributed run-tim
 e called Dragon specifically designed to manage dynamic processes, memory,
  and data on HPC systems. MLDocking automates the identification of top ca
 ndidates by executing all workflow components concurrently, efficiently di
 stributing tasks across CPU and GPU resources available on current heterog
 eneous HPC systems. Moreover, it limits the use of the file system by perf
 orming all data sharing operations through an in-memory distributed dictio
 nary that features local memory or fast RDMA transfers across the system’s
  interconnect. The talk will cover results obtained scaling the workflow o
 n the Aurora supercomputer and lessons learned in managing large datasets 
 for in-situ workflows.\n\nDomain: Chemistry and Materials, Climate, Weathe
 r, and Earth Sciences, Computational Methods and Applied Mathematics\n\nSe
 ssion Chairs: Riccardo Balin (Argonne National Laboratory) and Alessandro 
 Rigazzi (HPE)\n\n
END:VEVENT
END:VCALENDAR