BEGIN:VCALENDAR VERSION:2.0 PRODID:-//jEvents 2.0 for Joomla//EN CALSCALE:GREGORIAN METHOD:PUBLISH BEGIN:VEVENT UID:087ae01df3fb71123175ad9b2999ab19 CATEGORIES:Colloquium CREATED:20200114T195239 SUMMARY:Professor Pratyush Tiwary, University of Maryland LOCATION:CCB Auditorium (1303) DESCRIPTION:
"From atoms to mechanis ms: with a little help from statistical physics and artificial intelligence "
The ability to rapidly learn f
rom high-dimensional data to make reliable predictions about the future of
a given system is crucial in many contexts. This could be a fly avoiding pr
edators, or the retina processing terabytes of data almost instantaneously
to guide complex human actions. In this work we draw parallels between such
tasks, and the efficient sampling of complex molecules with hundreds of th
ousands of atoms. Such sampling is critical for predictive computer simulat
ions in condensed matter physics and biophysics, including but not limited
to problems such as crystal nucleation and drug unbinding. For this we use
the Predictive Information Bottleneck (PIB) framework developed and used fo
r the first two classes of problems, and re-formulate it for the sampling o
f biomolecular structure and dynamics, especially when plagued with rare ev
ents, and with minimum assumptions on the physics of the system [1-2]. Our
method considers a given biomolecular trajectory expressed in terms of orde
r parameters or basis functions, and uses a deep neural network to learn th
e minimally complex yet most predictive aspects of this trajectory, viz the
PIB. This information is used to perform iterative rounds of biased simula
tions that enhance the sampling along the PIB to gradually improve its accu
racy, directly obtaining associated thermodynamic and kinetic information.
We demonstrate the method on different test-pieces, where we calculate
the dissociation pathway and timescales slower than milliseconds. These&nb
sp;include ligand dissociation from the protein lysozyme and from flexible
RNA.
1. Tiwary and Berne, PNAS 2016
2. Wang, Ribeiro and Ti
wary, Nature Commun. 2019
~Coffee/tea will be served prior to lecture~
X-ALT-DESC;FMTTYPE=text/html:"From atoms to mechanisms: with a little help from statistical physics an d artificial intelligence"
The a
bility to rapidly learn from high-dimensional data to make reliable predict
ions about the future of a given system is crucial in many contexts. This c
ould be a fly avoiding predators, or the retina processing terabytes of dat
a almost instantaneously to guide complex human actions. In this work we dr
aw parallels between such tasks, and the efficient sampling of complex mole
cules with hundreds of thousands of atoms. Such sampling is critical for pr
edictive computer simulations in condensed matter physics and biophysics, i
ncluding but not limited to problems such as crystal nucleation and drug un
binding. For this we use the Predictive Information Bottleneck (PIB) framew
ork developed and used for the first two classes of problems, and re-formul
ate it for the sampling of biomolecular structure and dynamics, especially
when plagued with rare events, and with minimum assumptions on the physics
of the system [1-2]. Our method considers a given biomolecular trajectory e
xpressed in terms of order parameters or basis functions, and uses a deep n
eural network to learn the minimally complex yet most predictive aspects of
this trajectory, viz the PIB. This information is used to perform iterativ
e rounds of biased simulations that enhance the sampling along the PIB to g
radually improve its accuracy, directly obtaining associated thermodynamic
and kinetic information. We demonstrate the method on different test-pieces
, where we calculate the dissociation pathway and timescales slower th
an milliseconds. These include ligand dissociation from the protein ly
sozyme and from flexible RNA.
1. Tiwary and Berne, PNAS 2016
2. Wang, Ribeiro and Tiwary, Nature Commun. 2019
~ Coffee/tea will be served prior to lecture~
X-EXTRAINFO:Hosted by Professor Rick Remsing DTSTAMP:20240328T095537 DTSTART:20200303T160000 DTEND:20200303T170000 SEQUENCE:0 TRANSP:OPAQUE END:VEVENT END:VCALENDAR