Supercharged Protein Analysis in the Era of Accurate Structure Prediction

Martin Steinegger

/Seoul [KR]

Protein analysis has been transformed by machine-learning methods, with highly accurate structure prediction tools like AlphaFold2 and ESMFold leading the way. These methods have generated an unprecedented number of publicly available protein structures, with the AlphaFold database and ESMatlas now containing over 214 and 620 million predicted structures, respectively. To utilize this wealth of structural data, we have developed advanced tools like Foldseek, Foldseek-multimer and FoldMason to efficiently search and analyze these massive datasets. Additionally, our BFVD database significantly improves viral protein structure predictions by leveraging homology searches across petabases of sequencing data. This expanding landscape of structural information is revolutionizing genomic and proteomic annotations. In this talk, I will explore how these new tools and resources are enabling researchers to uncover novel biological insights and accelerate discovery across diverse fields of biology.

Go back