Coarse-Grained Force Field Optimisation: A Gaussian Process Regression Approach

Navarčíková, Petra (2024) Coarse-Grained Force Field Optimisation: A Gaussian Process Regression Approach. Bachelor's Thesis, Physics.

Preview	Text bPHYS2024NavarcikovaPetra.pdf Download (14MB) | Preview
	Text toestemming_ Petra Navarčíková _ degree programme_ Physics.pdf Restricted to Registered users only Download (131kB)

Abstract

Coarse-grained molecular dynamics simulations offer insight into fundamental biological processes which are challenging to investigate via all-atom simulations or complex experiments. Liquid-liquid phase separation drives formation of compartments, distinct chemical environments inside the cell, which directly affect cellular function and disease formation. In this work, a one-bead-per-amino-acid (1BPA) model is used to minimise the discrepancy between experimental and calculated radius of gyration (or Stokes radius) for a molecular data set comprising of 189 intrinsically disordered proteins (IDPs). The 1BPA force field is optimised via a supervised machine learning algorithm; Gaussian Process Regression. The GPR model predictions did not match molecular dynamics observations due to the small number of train data points and error metric definition in the target variable. Predictions were encompassed in the GPR confidence interval, which was relatively large due to under-fitting. Nevertheless, the newly determined 1BPA V3.0 shows considerable improvement compared to the previous 1BPA variants, with significant emphasis on aromatic amino acid interactions.

Item Type:	Thesis (Bachelor's Thesis)
Supervisor name:	Onck, P.R. and Giuntoli, A.
Degree programme:	Physics
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	08 Jul 2024 10:43
Last Modified:	08 Jul 2024 10:43
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/33139

Actions (login required)

View Item