Scanning transcriptomes for nonlinear, domain-level similarities using hmSEEKR

Area of Science:

• Genomics
• Molecular Biology
• Bioinformatics

Background:

• Long noncoding RNAs (lncRNAs) are crucial for gene regulation but often lack linear sequence similarity, hindering functional comparisons.
• Existing methods struggle to identify related lncRNAs when sequence homology is absent.

Purpose of the Study:

• To introduce hmSEEKR, a k-mer-based hidden Markov model for detecting non-linear sequence similarities in transcriptomes.
• To enable the discovery of functionally related RNA domains without prior knowledge of their location.

Main Methods:

• Developed and applied hmSEEKR, a k-mer-based hidden Markov model, to scan transcriptomes for non-linear sequence similarities.
• Utilized individual lncRNA domains as search features and performed transcriptome-wide and combinatorial searches.

Main Results:

• hmSEEKR successfully identified RNA regions with non-linear sequence similarity and similar protein-binding profiles.
• Domains within XIST, NEAT1, and MALAT1 showed widespread similarities to other RNAs, while some were unique.
• Combinatorial searches recovered RNAs with sequential matches to core functional domains, exhibiting similar biological properties and protein-interaction networks.

Conclusions:

• hmSEEKR offers a sequence-informed method for discovering functional elements in noncoding transcriptomes by identifying non-linear sequence similarities.
• The tool facilitates the discovery of RNA domains with shared regulatory functions, such as transcription promotion and hnRNP binding.