Machine Learning Integrating Protein Structure, Sequence, and Dynamics to Predict the Enzyme Activity of Bovine Enterokinase Variants

Aug 4, 2026

Upload by Noah

Despite advances in computational protein science, protein dynamics that govern biological activity cannot be inferred from sequence information alone. A machine learning framework is proposed to integrate peptide sequence, protein structure, and protein dynamics descriptors, combining traditional sequence/structure features with molecular dynamics simulation data. The resulting pipeline improves prediction of variant function by modeling the effects of multiple point mutations on fold improvement of bovine enterokinase activity. The study supports using structural and dynamic data to guide protein engineering for industrial applications.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages14

Reading Time35 min

AccessFree to download

Abstract

Introduction

Protein engineering challenges

Biodescriptors and machine learning

Sequence-based models vs. dynamics

Molecular Dynamics and ML integration

This article is licensed under CC-BY 4.0 [pubs.acs.org/jcim](pubs.acs.org/jcim) Article Machine Learning Integrating Protein Structure, Sequence, and Dynamics to Predict the Enzyme Activity of Bovine Enterokinase Variants Niccolo Alberto Elia Venanzi, Andrea Basciu, Attilio Vittorio Vargiu, Alexandros Kiparissides, Paul A. Dalby, and Duygu Dikicioglu * Cite This: [https://doi.org/10.1021/acs.jcim.3c00999](https://doi.org/10.1021/acs.jcim.3c00999) Read Online Downloaded via UNIV COLLEGE LONDON on February 26, 2024 at 13:57:38 (UTC) . See [https://pubs.acs.org/sharingguidelines](https://pubs.acs.org/sharingguidelines) for options on how to legitimately share published articles. ACCESS Metrics & More Article Recommendations *sı Supporting Information ABSTRACT: Despite recent advances in computational protein science, the dynamic behavior of proteins, which directly governs their biological activity, cannot be gleaned from sequence information alone. To overcome this challenge, we propose a framework that integrates the peptide sequence, protein structure, and protein dynamics descriptors into machine learning algorithms to enhance their predictive capabilities and achieve improved prediction of the protein variant function. The resulting machine learning pipeline integrates traditional sequence and structure information with molecular dynamics simulation data to predict the effects of multiple point mutations on the fold improvement of the activity of bovine enterokinase variants. This study highlights how the combination of structural and dynamic data can provide predictive insights into protein functionality and address protein engineering challenges in industrial contexts. ■ INTRODUCTION Proteins are essential, powerful machines in biology and consequently find a wide range of application areas in biotechnology, including the manufacturing of targeted therapies. However, their development as a functional product is expensive, time-consuming, and frequently yields unsuccessful results.1 The estimated average cost of bringing a new protein-based therapy to market is between $1 and $3 billion, and the success rate of clinical trials is below 10%. To overcome these obstacles, researchers are investigating new methods for expediting the engineering of proteins with enhanced functionality and manufacturability. Protein engineering entails making precise alterations to the original sequence of a protein to identify variants with desirable properties while minimizing interference with its function. Thus, protein engineering has revolutionized the production and use of protein-based products.2 However, due to the vast number of possible amino acid combinations, exhaustive experimental exploration of the landscape of protein fitness remains nearly impossible.3−6 Theoretically, the scope of mutations could be restricted to include only the ostensibly significant fragments of the protein such as the binding sites. However, this remains a heuristic solution applicable to a limited number of cases, as the majority of protein properties depend on the entire sequence and structural conformation, not just a few amino acids, due to the presence of epistatic effects.7,8 © XXXX The Authors. Published by American Chemical Society Biodescriptors are quantitative characteristics that shed light on a protein’s chemistry and structure. Algorithms can use the information contained in biodescriptors to predict the effects of amino acid substitutions on the properties of proteins. Recently, machine learning (ML) techniques have been applied to the classification of proteins and the prediction of the stability of protein−ligand complexes.9−13 Nevertheless, biodescriptors, which incorporate function-related properties associated with the macromolecules, were not employed in the investigation of the role of mutations introduced into the protein sequence on protein performance. 14−16 Unsupervised ML methods mostly use

FAQs

How is the method evaluated in the study?

The pipeline predicts the effects of multiple point mutations on the fold improvement of bovine enterokinase activity. This targets how mutations alter protein performance beyond static structural changes.

Aug 4, 2026

Upload by Noah