Dataset Information

Biomolecular simulation based machine learning models accurately predict sites of tolerability to the unnatural amino acid acridonylalanine.

ABSTRACT: The incorporation of unnatural amino acids (Uaas) has provided an avenue for novel chemistries to be explored in biological systems. However, the successful application of Uaas is often hampered by site-specific impacts on protein yield and solubility. Although previous efforts to identify features which accurately capture these site-specific effects have been unsuccessful, we have developed a set of novel Rosetta Custom Score Functions and alternative Empirical Score Functions that accurately predict the effects of acridon-2-yl-alanine (Acd) incorporation on protein yield and solubility. Acd-containing mutants were simulated in PyRosetta, and machine learning (ML) was performed using either the decomposed values of the Rosetta energy function, or changes in residue contacts and bioinformatics. Using these feature sets, which represent Rosetta score function specific and bioinformatics-derived terms, ML models were trained to predict highly abstract experimental parameters such as mutant protein yield and solubility and displayed robust performance on well-balanced holdouts. Model feature importance analyses demonstrated that terms corresponding to hydrophobic interactions, desolvation, and amino acid angle preferences played a pivotal role in predicting tolerance of mutation to Acd. Overall, this work provides evidence that the application of ML to features extracted from simulated structural models allow for the accurate prediction of diverse and abstract biological phenomena, beyond the predictivity of traditional modeling and simulation approaches.

SUBMITTER: Giannakoulias S

PROVIDER: S-EPMC8443755 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:Upon infection of a new host, human immunodeficiency virus (HIV) replicates in the mucosal tissues and is generally undetectable in circulation for 1-2 weeks post-infection. Several interventions against HIV including vaccines and antiretroviral prophylaxis target virus replication at this earliest stage of infection. Mathematical models have been used to understand how HIV spreads from mucosal tissues systemically and what impact vaccination and/or antiretroviral prophylaxis has on viral eradication. Because predictions of such models have been rarely compared to experimental data, it remains unclear which processes included in these models are critical for predicting early HIV dynamics. Here we modified the "standard" mathematical model of HIV infection to include two populations of infected cells: cells that are actively producing the virus and cells that are transitioning into virus production mode. We evaluated the effects of several poorly known parameters on infection outcomes in this model and compared model predictions to experimental data on infection of non-human primates with variable doses of simian immunodifficiency virus (SIV). First, we found that the mode of virus production by infected cells (budding vs. bursting) has a minimal impact on the early virus dynamics for a wide range of model parameters, as long as the parameters are constrained to provide the observed rate of SIV load increase in the blood of infected animals. Interestingly and in contrast with previous results, we found that the bursting mode of virus production generally results in a higher probability of viral extinction than the budding mode of virus production. Second, this mathematical model was not able to accurately describe the change in experimentally determined probability of host infection with increasing viral doses. Third and finally, the model was also unable to accurately explain the decline in the time to virus detection with increasing viral dose. These results suggest that, in order to appropriately model early HIV/SIV dynamics, additional factors must be considered in the model development. These may include variability in monkey susceptibility to infection, within-host competition between different viruses for target cells at the initial site of virus replication in the mucosa, innate immune response, and possibly the inclusion of several different tissue compartments. The sobering news is that while an increase in model complexity is needed to explain the available experimental data, testing and rejection of more complex models may require more quantitative data than is currently available.

Dataset Information

Biomolecular simulation based machine learning models accurately predict sites of tolerability to the unnatural amino acid acridonylalanine.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets