BEGIN:VCALENDAR
VERSION:2.0
PRODID:icalendar-ruby
CALSCALE:GREGORIAN
BEGIN:VEVENT
DTSTAMP:20260627T063210Z
UID:cb86daac-2073-4bbc-a810-1a544b6fe5cf
DTSTART:20241008T170000Z
DTEND:20241008T170000Z
DESCRIPTION:This webinar is part of a series run by the ELIXIR 3D-BioInfo C
 ommunity. There is a complete list of webinars here.  \n			\n				 \n			
 \n		\n	\n\n\nHosts\n\n \n	\n		\n			Dr. Gonzalo Parra\n				Barcelona Super
 computing Center\n		\n	\n\n	\n		\n			Dr Neeladri Sen\n				University Coll
 ege London (UCL)\n		\n	\n\n\nProgramme\n\nPredicting protein conformationa
 l motions using energetic frustration analysis and AlphaFold2\n\nXingyue G
 uan (Department of Physics\, National Laboratory of Solid State Microstruc
 ture\, Nanjing University\, Nanjing 210093\, China\,\n\n\n\n\n	Proteins pe
 rform their biological functions through motion. Although high throughput 
 prediction of the three-dimensional static structures of proteins has prov
 ed feasible using deep-learning-based methods\, predicting the conformatio
 nal motions remains a challenge. Purely data-driven machine learning metho
 ds encounter difficulty for addressing such motions because available labo
 ratory data on conformational motions are still limited. In this work\, we
  develop a method for generating protein allosteric motions by integrating
  physical energy landscape information into deep-learning-based methods. W
 e show that local energetic frustration\, which represents a quantificatio
 n of the local features of the energy landscape governing protein alloster
 ic dynamics\, can be utilized to empower\n\n	\n		\n			 AlphaFold2 (AF2) t
 o predict protein conformational motions. Starting from ground state stati
 c structures\, this integrative method generates alternative structures as
  well as pathways of protein conformational motions\, using a progressive 
 enhancement of the energetic frustration features in the input multiple se
 quence alignment sequences. For a model protein adenylate kinase\, we show
  that the generated conformational motions are consistent with available e
 xperimental and molecular dynamics simulation data. Applying the method to
  another two proteins KaiB and ribose-binding protein\, which involve larg
 e-amplitude conformational changes\, can also successfully generate the al
 ternative conformations. We also show how to extract overall features of t
 he AF2 energy landscape topography\, which has been considered by many to 
 be black box. Incorporating physical knowledge into deep learning-based st
 ructure prediction algorithms provides a useful strategy to address the ch
 allenges of dynamic structure prediction of allosteric proteins.\n\n			Pre
 dicting protein fold switching with AlphaFold and other approaches\n\n			D
 r. Lauren Porter(National Institute of Health\, USA) \n\n			\n				Recent 
 work suggests that AlphaFold (AF)–a deep learning-based model that can a
 ccurately infer protein structure from sequence–may discern important fe
 atures of folded protein energy landscapes\, defined by the diversity and 
 frequency of different conformations in the folded state. Here\, we test t
 he limits of its predictive power on fold-switching proteins\, which assum
 e two structures with regions of distinct secondary and/or tertiary struct
 ure.\n\n				We find that (1) AF is a weak predictor of fold switching and 
 (2) some of its successes result from memorization of training-set structu
 res rather than learned protein energetics. Combining &gt\;280\,000 models
  from several implementations of AF2 and AF3\, a 35% success rate was achi
 eved for fold switchers likely in AF’s training sets. AF2’s confidence
  metrics selected against models consistent with experimentally determined
  fold-switching structures and failed to discriminate between low and high
  energy conformations.\n\n				Further\, AF captured only one out of seven 
 experimentally confirmed fold switchers outside of its training sets despi
 te extensive sampling of an additional ~280\,000 models. Several lines of 
 evidence suggest that AF2 has memorized structural information during trai
 ning\, and AF3 misassigns coevolutionary restraints. These limitations con
 strain the scope of successful predictions and suggest approaches to impro
 ve predictive robustness in the future.\n\n				You can find previous webin
 ars from the 3D-BioInfo Community on the Community webinars page.
SUMMARY:3DSig 2024: Conformational Diversity
URL;VALUE=URI:https://www.elixir-europe.org/events/3dsig-2024-conformationa
 l-diversity
END:VEVENT
END:VCALENDAR