🚀 @SBERLOGABIG online seminar on bioinformatics and machine learning:
👨🔬 Laurentiu Spiridon «Generalized Coordinates Hamiltonian Monte Carlo Gibbs Sampling Based on Robot Mechanics»
⌚️ Пятница 25 марта, 19.00 по Москве
(Доклад - на английском)
Biomolecular structures sampling is still a difficult task due to the sizes of the systems involved as for example the in the case of Sars-Cov-2 Spike protein which has ~70000 atoms. Therefore developing better simulation methods that can generate large pools of uncorrelated samples is currently of central focus in structural bioinformatics. One of the main applications is creating ensemble of complex conformations for free energy calculations which can dramatically improve results especially if large conformational changes are possible.
Macromolecules conformations distributions, recovered by regular molecular dynamics and the majority of Markov Chain Monte Carlo methods, are either distorted or too narrow respectively. Recently, we successfully applied robot mechanics algorithms to these cases with considerable increase in efficiency (Spiridon et. al, 2020) owing to the integration of Gibbs sampling in Generalized Coordinates Hamiltonian Monte Carlo (GCHMC). This approach allows the use of large rigid bodies as constraints to complement either soft degrees of freedom such as torsions or key regions that induce high amplitude motions.
More preciselly in Robosample, Gibbs blocks are were defined as degrees of freedom mapped onto joints between rigid bodies. Kinematics and dynamics algorithms borrowed from robot mechanics render the method as fast as regular molecular simulation programs. Ergodicity is achieved by including a block for the joint probability as Cartesian HMC.
Molecular conformational sampling based on rigid body dynamics presents itself as a promising
method to collect macromolecules‘ conformations being independent from other enhanced
sampling methods which can be invoked in parallel.
References:
Spiridon L, Minh DDL, "Hamiltonian Monte Carlo with Constrained Molecular Dynamics as
Gibbs Sampling" J Chem Theory Comput. 2017, 13, 4649-4659.
Spiridon L, Şulea TA, Minh DDL, Petrescu AJ. Robosample: “Robosample: A rigid-body
molecular simulation program based on robot mechanics.” Biochim Biophys Acta Gen Subj.
2020 Aug;1864(8):129616.
О докладчике: Laurentiu Spiridon is a researcher in the Institute of Biochemistry of the Romanian Academy. Laurentiu is currently working in Systems Biology of Aging, Bioinformatics & Structural Biochemistry in the Bioinformatics & Structural Biochemistry
Ссылка на зум будет доступна в этом канале: https://t.me/sberlogabig перед началом доклада.
Подписывайтесь на канал @sberlogabig ! Интересные материалы и увлекательные доклады.👍
👨🔬 Laurentiu Spiridon «Generalized Coordinates Hamiltonian Monte Carlo Gibbs Sampling Based on Robot Mechanics»
⌚️ Пятница 25 марта, 19.00 по Москве
(Доклад - на английском)
Biomolecular structures sampling is still a difficult task due to the sizes of the systems involved as for example the in the case of Sars-Cov-2 Spike protein which has ~70000 atoms. Therefore developing better simulation methods that can generate large pools of uncorrelated samples is currently of central focus in structural bioinformatics. One of the main applications is creating ensemble of complex conformations for free energy calculations which can dramatically improve results especially if large conformational changes are possible.
Macromolecules conformations distributions, recovered by regular molecular dynamics and the majority of Markov Chain Monte Carlo methods, are either distorted or too narrow respectively. Recently, we successfully applied robot mechanics algorithms to these cases with considerable increase in efficiency (Spiridon et. al, 2020) owing to the integration of Gibbs sampling in Generalized Coordinates Hamiltonian Monte Carlo (GCHMC). This approach allows the use of large rigid bodies as constraints to complement either soft degrees of freedom such as torsions or key regions that induce high amplitude motions.
More preciselly in Robosample, Gibbs blocks are were defined as degrees of freedom mapped onto joints between rigid bodies. Kinematics and dynamics algorithms borrowed from robot mechanics render the method as fast as regular molecular simulation programs. Ergodicity is achieved by including a block for the joint probability as Cartesian HMC.
Molecular conformational sampling based on rigid body dynamics presents itself as a promising
method to collect macromolecules‘ conformations being independent from other enhanced
sampling methods which can be invoked in parallel.
References:
Spiridon L, Minh DDL, "Hamiltonian Monte Carlo with Constrained Molecular Dynamics as
Gibbs Sampling" J Chem Theory Comput. 2017, 13, 4649-4659.
Spiridon L, Şulea TA, Minh DDL, Petrescu AJ. Robosample: “Robosample: A rigid-body
molecular simulation program based on robot mechanics.” Biochim Biophys Acta Gen Subj.
2020 Aug;1864(8):129616.
О докладчике: Laurentiu Spiridon is a researcher in the Institute of Biochemistry of the Romanian Academy. Laurentiu is currently working in Systems Biology of Aging, Bioinformatics & Structural Biochemistry in the Bioinformatics & Structural Biochemistry
Ссылка на зум будет доступна в этом канале: https://t.me/sberlogabig перед началом доклада.
Подписывайтесь на канал @sberlogabig ! Интересные материалы и увлекательные доклады.👍