Predicting isoforms functional importance with TRIFID

We are pleased to announce TRIFID, a Machine Learning-based method for predicting isoform functional importance.

The advent of long-read sequencing means that the number of annotated transcripts in reference databases will increase. Therefore, it is crucial to understand which protein isoforms are biologically relevant and which are not. TRIFID overcomes this challenge, harnessing proteomics evidence as a proxy for functionality.  To do so, the algorithm evaluates over 40 predictive features for both principal and alternative isoforms divided into 5 categories: annotation, evolution, expression, structure, and splicing. The model has been trained on isoforms detected in large-scale proteomics analyses to distinguish these biologically important splice isoforms with high confidence.

In particular, we hope TRIFID will be a particularly valuable tool to help understand the pathogenic effects of mutations on splice variants.

TRIFID has been published at NAR Genomics and Bioinformatics and is available here.