BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Denver
X-LIC-LOCATION:America/Denver
BEGIN:DAYLIGHT
TZOFFSETFROM:-0700
TZOFFSETTO:-0600
TZNAME:MDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0600
TZOFFSETTO:-0700
TZNAME:MST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20200129T163603Z
LOCATION:502-503-504
DTSTART;TZID=America/Denver:20191117T103000
DTEND;TZID=America/Denver:20191117T105500
UID:submissions.supercomputing.org_SC19_sess101_ws_dls107@linklings.com
SUMMARY:DeepDriveMD: Deep-Learning Driven Adaptive Molecular Simulations f
 or Protein Folding
DESCRIPTION:Workshop\n\nDeepDriveMD: Deep-Learning Driven Adaptive Molecul
 ar Simulations for Protein Folding\n\nLee, Turilli, Jha, Bhowmik, Ma...\n\
 nSimulations of biological macromolecules are important in understanding t
 he physical basis of complex processes such as protein folding. However, e
 ven with increasing computational capacity and specialized architectures, 
 the ability to simulate protein folding at atomistic scales still remains 
 challenging. This stems from the dual aspects of high dimensionality of pr
 otein conformational landscapes, and the inability of atomistic molecular 
 dynamics (MD) simulations to sufficiently sample these landscapes to obser
 ve folding events. Machine learning/deep learning (ML/DL) techniques, when
  combined with atomistic MD  simulations offer the opportunity to potentia
 lly overcome these limitations by: (1) effectively reducing the dimensiona
 lity of MD simulations to automatically build latent representations that 
 correspond to biophysically relevant reaction coordinates (RCs), and (2) d
 riving MD simulations to automatically sample potentially novel conformati
 onal states based on these RCs. We examine how coupling DL approaches with
  MD simulations can lead to effective approaches to fold small proteins on
  supercomputers. In particular, we study the computational costs and effec
 tiveness of scaling DL-coupled MD workflows implemented using RADICAL-Cybe
 rtools in folding two prototypical systems, namely Fs-peptide and the fast
 -folding variant of the villin head piece protein. We demonstrate that a D
 L-coupled MD workflow is able to effectively learn latent representations 
 and drive adaptive simulations. Compared to traditional MD-based approache
 s, our approach achieves an effective performance gain in sampling the fol
 ded states by at least 2.3x. Together, our study provides quantitative bas
 is to understand how coupling DL approaches to MD simulations, can lead to
  effective performance gains and reduced times to solution on supercomputi
 ng resources.\n\nTag: Workshop Reg Pass, Deep Learning, Scientific Computi
 ng\n\nRegistration Category: Workshop Reg Pass, Deep Learning, Scientific 
 Computing
URL:https://sc19.supercomputing.org/presentation/?id=ws_dls107&sess=sess10
 1
END:VEVENT
END:VCALENDAR