Dr. David (Dudu) Burstein

In our lab we use LLMs, deep learning, feature-based machine learning to better understand key interaction mechanisms within microbial communities and to promote their application in biotechnology and medicine. By combining cutting-edge machine learning approaches on large genomic and metagenomic datasets, we decipher functions in the nexus of microbial interactions, such as CRISPR-Cas, anti-CRISPRs, bacterial pathogenesis systems, and antibiotic resistance genes.

Recently, one of our main interests is the integration of techniques adopted from natural language processing (NLP) to understand "gene syntax and semantics". In our initial work on the topic, we treated genes as words and genomic fragments as sentences. Leveraging NLP methodologies, we successfully predicted the function of previously uncharacterized genes with high accuracy. We are now integrating transformer-based large language models, harnessing state-of-the-art NLP models to better understand different aspects of microbial systems through this novel lens.