Alessandra Carbone
Laboratoire de Biologie Computationnelle et Quantitative, Sorbonne Université-CNRS, Paris, France
Accurately annotating missense variants is pivotal in the fight against rare diseases. Despite advancements, such as the breakthrough AlphaMissense model featured in Science on September 23, 2023, clinicians continue to face the challenge of distinguishing pathogenic rare variants from rare benign polymorphisms—a distinction critical for understanding the pathogenic potential of rare missense variants. In response, we released PRESCOTT in January 2024 (prescott.lcqb.upmc.fr), a computational model providing exhaustive mutational landscapes for over 19,000 human proteins. This tool not only identifies mutation-prone regions in proteins, but also scores missense variants to aid pathologists in evaluating the pathogenic potential of specific mutations. PRESCOTT has demonstrated superior performance compared to AlphaMissense in its ability to predict pathogenic rare variants, especially in complex scenarios involving gain-of-function mutations in proteins related to autoinflammatory diseases. Both AlphaMissense and PRESCOTT leverage multiscale biological data, including sequences, structures, and population allele frequencies. I will discuss the problem, our model, and its significant impact on genomic medicine.
