De Novo Clustering of Long-Read Transcriptome Data Using a Greedy, Quality Value-Based Algorithm.

Area of Science:

• Genomics
• Bioinformatics
• Transcriptomics

Background:

• Long-read sequencing technologies like PacBio Iso-Seq and Oxford Nanopore enable detailed studies of transcript isoforms.
• Current algorithms struggle with accurate and scalable clustering of long reads, limiting the full potential of these technologies.
• Clustering reads by gene family is a critical bottleneck in de novo transcript reconstruction.

Purpose of the Study:

• To develop a novel, scalable, and accurate algorithm for clustering long-read sequencing data.
• To address the limitations of existing methods in handling complex isoform landscapes.
• To improve the accuracy and efficiency of de novo transcript reconstruction.

Main Methods:

• Developed isONclust, a greedy clustering algorithm that incorporates read quality values.
• Tested isONclust on simulated and biological datasets across various organisms, technologies, and read depths.
• Compared isONclust performance against existing clustering approaches.

Main Results:

• isONclust demonstrates substantial improvements in accuracy compared to previous methods.
• The algorithm offers enhanced scalability for processing large long-read sequencing datasets.
• Performance was validated across diverse biological contexts and sequencing platforms.

Conclusions:

• isONclust effectively overcomes key limitations in long-read transcript analysis.
• The algorithm enhances the utility of PacBio Iso-Seq and Oxford Nanopore data for transcriptomic studies.
• isONclust represents a significant advancement for scalable and accurate isoform reconstruction.