Key points from article :
Alchemab Therapeutics has unveiled a ground-breaking machine learning model called AntiBERTa, which analyses antibody sequences to better understand their structure and function. Published in the journal Patterns, AntiBERTa is a 12-layer transformer neural network that provides detailed insights into antibody sequences, helping to identify biologically relevant patterns. This model can predict antibody binding sites and learn critical characteristics of B cell receptors (BCRs), such as their mutational load, activation level, and structure.
The study revealed that AntiBERTa can distinguish between naïve and memory B cells and identify pairs of amino acid positions in the BCR sequence that interact in three-dimensional space. These findings suggest that the model captures essential information about BCRs, such as their immunogenicity and function. This technology is already aiding Alchemab’s antibody drug discovery process by identifying convergent antibodies linked to disease resilience.
Dr. Jane Osbourn, Alchemab’s CSO, highlighted that AntiBERTa will accelerate the discovery of novel targets for diseases like cancer and neurodegenerative disorders. Alchemab’s platform uses deep sequencing and computational analysis of patient antibody repertoires to uncover new therapeutic targets. By analysing naturally optimized antibodies, the company aims to develop innovative treatments and improve patient stratification in the fields of oncology and immunotherapy.
