MARS and RNAcmap3: The Master Database of All Possible RNA Sequences Integrated with RNAcmap for RNA Homology Search

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Protein structure prediction tools like AlphaFold2 leverage co-evolutionary data from large sequence databases.
• Existing nucleotide databases lack consolidation, hindering comprehensive homology searches for RNA sequences.
• Non-coding RNAs (ncRNAs) play crucial roles, but their functional and structural analysis is often limited by database accessibility.

Purpose of the Study:

• To create a consolidated and significantly larger database for RNA sequence homology searches.
• To develop a new strategy for improved homology searching and multiple sequence alignment (MSA) generation for ncRNAs.
• To enhance the accuracy and sensitivity of structural and functional inference for ncRNAs.

Main Methods:

• Integrated diverse RNA sequence data from RNAcentral, MG-RAST, GWH, MGnify, and NCBI's nt database.
• Developed the Master database of All possible RNA sequences (MARS), achieving a 20-fold increase over NCBI's nt database.
• Implemented a novel split-search strategy for enhanced homology detection and utilized the RNAcmap tool for automatic alignment.

Main Results:

• MARS is substantially larger than existing nucleotide databases, enabling deeper homology searches.
• The new dataset and split-search strategy significantly outperform current state-of-the-art homology search techniques.
• Generated more accurate and sensitive multiple sequence alignments (MSAs) compared to manually curated Rfam alignments for many structured RNAs.

Conclusions:

• MARS provides a powerful resource for advancing ncRNA research.
• The MARS database coupled with RNAcmap facilitates improved structural and functional inference of ncRNAs.
• This resource is valuable for developing advanced RNA language models based on MSAs.