PredIDR3: A new output-encoding scheme and abundant negative source provide more information for deep learning-based protein intrinsic disorder prediction
Many computational methods to predict intrinsic disordered regions (IDRs) in proteins have been developed and their performances are blindly evaluated in community-driven assessment, Critical Assessment of protein Intrinsic Disorder (CAID). In this study, we developed PredIDR3 series, an updated version of PredIDR2 tested in CAID3 to accurately predict IDRs from protein sequences. It includes two…
The PredIDR3 series represents a significant advancement in the field of predicting intrinsic disordered regions (IDRs) in proteins. Building upon the foundation laid by its predecessor, PredIDR2, which has been evaluated in the Critical Assessment of Protein Intrinsic Disorder (CAID), PredIDR3 introduces a more accurate and comprehensive approach to IDR prediction.
By incorporating an ensemble method, PredIDR3 demonstrates a marked improvement in performance, with an AUC_ROC score of 0.953 compared to PredIDR2's 0.936, on the Disorder-PDB dataset of CAID3. This enhancement is primarily attributed to the utilization of more information for intrinsic disorder prediction, achieved through the implementation of a novel output-encoding scheme that allows for a larger sliding window size of 91.
Furthermore, PredIDR3 introduces the extraction of negative samples from non-IDRs within the PDB and DisProt databases for training, effectively leveraging additional information to bolster the prediction capabilities. The performance of PredIDR3 series is found to be comparable to the leading methods in CAID3 across all evaluation criteria.
Researchers and practitioners interested in exploring this innovative approach can access PredIDR3 through the CAID Prediction Portal at https://caid.idpcentral.org/portal or download it as a Singularity container from https://biocomputingup.it/shared/caid-predictors/.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
