Biology subjects

Mafofo, J.

Publications and source records attributed to Mafofo, J..

2 recordsLinked to original sources

bioRxiv2022

CRAM compression: practical across-technologies considerations for large-scale sequencing projects

CRAM is an efficient format to store high-throughput sequencing data and it has been widely adopted. We thus plan to use CRAM for the Emirati Genome Program, which aims to sequence the genomes of ~1 million nationals in the United Arab Emirates using short- and long-read sequencing technologies (Illumina, MGI and Oxford Nanopore Sequencing). We conducted a pilot study on the three technologies before start using CRAM at scale. We found CRAM achieved 40-70% compression depending on the sequencing platform. As expected, CRAM compression was data lossless and did not alter variant calls. In our cloud, we observed compression speeds 0.7-1.4 GB per minute, varying on the sequencing platform too. This translates into ~1-2 hours using a single CPU to compress a ~30X human whole-genome sequencing sample. Despite its wide use, we found little publicly available information about CRAM compression rate, speed, losslessness and parallelization, especially across many sequencing platforms. This work will have direct application for Emirati Genome Program and provide practical considerations for other large-scale sequencing efforts.

bioinformatics↗

bioRxiv2020

Fast-Bonito: A Faster Basecaller for Nanopore Sequencing

Oxford Nanopore Technologies (ONT) is a promising sequencing technology that could generate relatively longer sequencing reads compared to the next generation sequencing (NGS) technology. The base calling process is very important for TGS. It translates the original electrical signals from the sequencer to the nucleotide sequence. By doing that, the base calling could significantly influence the accuracy of downstream analysis. Bonito is a recently developed basecaller based on deep neuron network, the neuron network architecture of which is composed of a single convolutional layer followed by three stacked bidirectional GRU layers. Although Bonito achieved the state-of-the-art accuracy, its speed is so slow that it is not likely to be used in production. We therefore implement Fast-Bonito, which introduces systematic optimization to speed up Bonito. Fast-Bonito archives 53.8% faster than the original version on NVIDIA V100 and could be further speed up by HUAWEI Ascend 910 NPU, achieving 565% faster than the original version. The accuracy of Fast-Bonito is also slightly higher than the original Bonito.

bioinformatics↗