A Machine Learning Approach for Predicting Antibody Properties

Oche A. Egaji, Seamus Ballard-Smith, Ikram Asghar, Mark Griffiths

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This paper used an amino acid location-based sequence encoding as a feature extraction techniques to identify single chains antibody molecules that bind to B-lymphocyte stimulator (BLyS) antigen. The data were manually derived from the European patent (EP2275449B1) text. The dataset was cleaned and made suitable for the machine learning models. The accuracy, precision and recall achieved across individual descriptors (Membrane and Soluble) for Logistic regression, KNN, KSVM, and Random Forest Tree was above 80%. However, it was much lower for the Naïve Bayes except for the precision score. The promising accuracy value achieved from such a minimal dataset has significant implications for the drug discovery process - this includes considerable savings in time and resources.
Original languageEnglish
Title of host publicationProceedings of the 10th International Conference on Information Communication and Management
Pages20–24
Number of pages5
DOIs
Publication statusPublished - 24 Sept 2020
Externally publishedYes
Event10th International Conference on Information Communication and Management - Paris, France
Duration: 12 Aug 202014 Aug 2020
Conference number: 10

Conference

Conference10th International Conference on Information Communication and Management
Country/TerritoryFrance
CityParis
Period12/08/2014/08/20

Fingerprint

Dive into the research topics of 'A Machine Learning Approach for Predicting Antibody Properties'. Together they form a unique fingerprint.

Cite this