Sirocco Research Labs

Algorithms for Multiscale Characterization of Equilibrium Protein Motions

Find­ings by Andrej Savol

Con­for­ma­tional motions rel­e­vant to pro­tein func­tion span an enor­mous range of tem­po­ral scales. While the appro­pri­ate method for under­stand­ing these fluc­tu­a­tions may be spe­cific to a par­tic­u­lar timescale and detail level, the char­ac­ter­i­za­tion of spa­tial and ener­getic tran­si­tions pro­vides our lab’s over­ar­ch­ing focus. To that aim, we have devel­oped advanced machine learn­ing tech­niques for extract­ing non-Gaussian and rare behav­ior in MD tra­jec­to­ries. Although out­comes of purely spa­tial fluc­tu­a­tions, QAA basis vec­tors seg­ment ubiquitin’s hier­ar­chi­cal inter­nal energy space with sur­pris­ing clar­ity, and we give insight for why pur­su­ing such rare events pro­vides use­ful bio­chem­i­cal infor­ma­tion, espe­cially when com­pared with estab­lished principal-component and full-(de)correlation based approaches.

Longer tem­po­ral win­dows and larger spa­tial motions are con­sid­ered with Struc­ture–from–FRET, an algo­rithm that answers the fol­low­ing ques­tion: given a time-evolving dis­tance con­straint between two pro­tein residues, what is the most likely tra­jec­tory of the entire pro­tein? Our Kalman-Filter based approach gen­er­ates ensem­bles of tra­jec­to­ries that agree with sin­gle mol­e­cule Forster Res­o­nance Energy Trans­fer data. We high­light its util­ity with Adeny­late Kinase, a model hinge pro­tein, and describe how this appar­ently under­de­ter­mined prob­lem is made tractable by incor­po­rat­ing stan­dard bio­chem­i­cal con­straints (bond-length, bond-angle) into the state-space update algorithm.

Published June 28th, 2010 by Jimmy Marble

Comments

• GSX-R750 guy

  July 4, 2010 at 4:33 am

  Where did you learn about this? Can you give me the source?
• Jimmy Marble

  July 7, 2010 at 12:40 pm

  Andrej Savol heads the Sirocco Research Labs’ Sci­ence Labs. He does his research at the Uni­ver­sity of Pittsburgh.