Pedro Barbosa, LASIGE’s PhD student, published a paper “Computational prediction of human deep intronic variation”, co-authored with Alcides Fonseca, LASIGE integrated member, among others, in Giga Science, a top-ranked journal.
The work is a collaborative effort between LASIGE and Instituto de Medicina Molecular João Lobo Antunes, conducted as part of Pedro Barbosa’s PhD project. This work features an effort to study non-coding regions of the human genome known as introns. Previously dismissed as “junk DNA”, introns are now recognized to play a crucial role in gene regulation, especially in RNA splicing, a process that joins coding DNA (exons) to create functional proteins.
This paper independently assessed various computational methods for predicting the impact of disease-causing genetic variants within deep intronic regions of the genome. Besides revealing specific deep-learning models that excel at the task, the authors were able to provide recommendations for applying these models in clinical practice. Additionally, they made an initial attempt to assess their explainability in relation to disease phenotypes. This extensive work involved a thorough process of data curation, the creation of benchmark datasets, the integration of 38 different models, and the establishment of a standardized benchmark framework (VETA) that streamlines the evaluation of models working with genetic variant data.
The paper is available here.