DeepCLIP: predicting the effect of mutations on protein-RNA binding with deep learning

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Nucleotide variants can alter protein-RNA binding, impacting critical cellular processes like splicing and transcript stability.
• Accurate prediction of protein-RNA binding is essential for understanding the functional consequences of sequence variations.
• The diverse recognition motifs of RNA-binding proteins and context-dependent binding present significant modeling challenges.

Purpose of the Study:

• To develop a novel method, DeepCLIP, for context-aware modeling and prediction of protein binding to RNA using only sequence data.
• To evaluate DeepCLIP's performance against existing methods for RNA-protein binding prediction.
• To demonstrate the utility of DeepCLIP in predicting functional outcomes of nucleotide variants and in therapeutic applications.

Main Methods:

• DeepCLIP utilizes exclusively sequence data as input for context-aware modeling of protein-RNA interactions.
• Performance evaluation involved comparing DeepCLIP predictions with existing RNA-protein binding modeling methods.
• Validation included correlating DeepCLIP predictions with functional outcomes from wet lab experiments.

Main Results:

• DeepCLIP demonstrates superior performance in modeling RNA-protein binding compared to existing methods.
• DeepCLIP predictions show a significant correlation with the functional effects of nucleotide variants observed in experimental studies.
• The study highlights DeepCLIP's capability in designing antisense oligonucleotides and uncovering tissue-specific regulatory mechanisms.

Conclusions:

• DeepCLIP represents a significant advancement in predicting RNA-protein binding and its functional impact.
• The method offers a powerful tool for variant effect prediction, therapeutic design, and understanding gene regulation.
• DeepCLIP is available as a freely accessible application and webtool.