CATHe2: Enhanced CATH Superfamily Detection Using ProstT5 and Structural Alphabets
MotivationThe CATH database is a free publicly available online resource that provides annotations about the evolutionary and structural relationships of protein domains. Due to the flux of protein structures coming mainly from the recent breakthrough of AlphaFold and therefore the non-feasibility of manual intervention, the CATH team recently developed an automatic CATH superfamily classifier called CATHe, that uses a feed-forward network classifier with protein Language Model (pLM) embeddings as input. Using the same dataset, in this paper, we present, CATHe2 that improves on CATHe by switching the old pLM ProtT5 for one of the most recent versions called ProstT5, and by introducing domain 3D information as input to the classifier, in the form of Structural Alphabet representation, namely 3Di sequence embeddings. Finally, CATHe2 implements a new version of the feed-forward network (FNN, i.e, non-recurrent neural network) classifier architecture, fine-tuned to perform at the CATH superfamily prediction task. ResultsThe best CATHe2 model reaches an accuracy of 92.2 {+/-} 0.7% with an F1 score of 82.3 {+/-} 1.3% which constitutes an improvement of 9.9% on the F1 score and 6.6% on the accuracy, from the previous CATHe version (85.6 {+/-} 0.4% accuracy and 72.4 {+/-} 0.7% F1 score) on its largest dataset (~ 1700 superfamilies). This model uses ProstT5 AA sequence and 3Di sequence embeddings as input to the classifier, but a simplified version requiring only AA sequences, already improves CATHes F1 score by 6.7 {+/-} 1.3% and accuracy by 6.6 {+/-} 0.7% on its largest dataset. Availability & ImplementationThe code is available on https://GitHub.com/Mouret-Orfeu/CATHe2. Datasets: https://doi.org/10.5281/zenodo.14534966 Contactorfeu.mouret.pro@outlook.fr, j.abbass@kingston.ac.uk