SPLASH2 provides ultra-efficient, scalable, and unsupervised discovery on raw sequencing reads

Area of Science:

• Bioinformatics
• Genomics
• Computational Biology

Background:

• Discovering sequence variations and biological insights from large-scale sequencing data is computationally challenging.
• Existing methods are often application-specific and require manual tuning.
• Unsupervised and reference-free approaches are needed for broad applicability.

Approach:

• Introduced SPLASH2, a fast and scalable implementation of the SPLASH algorithm.
• Utilizes an efficient k-mer counting approach for rapid analysis of massive datasets.
• Designed for unsupervised, reference-free discovery of regulated sequence variation.

Key Points:

• SPLASH2 enables rapid analysis across diverse sequencing technologies and biological contexts.
• Unveiled new, unannotated alternative splicing in cancer transcriptomes from the Cancer Cell Line Encyclopedia (CCLE).
• Demonstrated high precision and scalability by detecting BCR-ABL gene fusions and circRNAs.

Conclusions:

• SPLASH2 offers unparalleled scale and speed for analyzing massive RNA-seq data.
• The algorithm effectively uncovers novel biological insights without parameter tuning.
• SPLASH2 is a versatile tool for diverse RNA-seq detection tasks, including gene fusion and circRNA identification.