Biological drugs, such as monoclonal antibodies (mAbs) are receiving increasing attention, as their market size is expected to grow in the near future. This is due to mAbs being highly specific towards an antigen epitope, which is determined by the variable region (Fv). Due to their complex structure, therapeutic mAbs represent challenging molecules for use of computational methods. However, recent studies demonstrate that the complexity of sequential patterns found in proteins can be captured by different deep learning algorithms.<br/><br/>In this study the Fv region was used with different deep learning approaches to investigate if it is possible to capture the pattern relationship between the amino acid sequence and the target groups. Fv region for all the mAbs used were extracted from the Thera-SAbDAb database and then divided into two target groups, CD-antigen and other. The size of the other target group was adjusted to the size of the CD-antigen group in order to balance the data set. The sequences for the light and heavy chains were padded, concatenated and then one-hot encoded.<br/><br/>The models obtained show a prediction accuracy of around 0.7, without performing additional exhaustive approaches to improve the model performance.

Deep learning approaches to investigate sequence-target group mapping for monoclonal antibodies

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.