ReadChop: a high-performance demultiplexer for long-read sequencing data

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Long-read sequencing (LRS) offers extended read lengths but faces computational hurdles like high error rates and indel artifacts.
• Sample multiplexing is crucial for cost-efficiency in LRS, but existing demultiplexing tools struggle with flexibility for non-canonical designs or speed for complex barcodes.

Purpose of the Study:

• To develop a high-performance demultiplexer for long-read sequencing data.
• To address the limitations of existing tools in handling complex barcode designs and computational challenges.

Main Methods:

• Developed ReadChop, a demultiplexer implemented in Rust, utilizing Myers' bit-parallel algorithm for indel-rich error profiles.
• Employed a streaming architecture for low memory footprint and efficient processing of large datasets.
• Benchmarked ReadChop on simulated and empirical datasets, including ultra-high multiplexing (13,824-plex) and SARS-CoV-2 amplicons.

Main Results:

• Achieved over 99.99% classification precision on simulated and empirical data.
• Demonstrated efficient filtering of in silico chimeras with a 0.1% miss rate.
• Showcased linear computational scalability on ultra-long templates (up to 100 kb) with memory usage consistently below 200 MB.
• Exceeded performance benchmarks, being >6x faster than Dorado, >2x faster than Nanoplexer, and >30x faster than Cutadapt.

Conclusions:

• ReadChop provides a flexible, robust, and high-performance solution for demultiplexing massive long-read sequencing datasets.
• The tool effectively handles non-canonical experimental designs and complex barcode structures.
• ReadChop significantly improves processing speed and efficiency compared to existing demultiplexing software.