The enduring advantages of the SLOW5 file format for raw nanopore sequencing data.

Abstract

Nanopore sequencing is a widespread and important method in genomics science. The raw electrical current signal data from a typical nanopore sequencing experiment are large and complex. This can be stored in 2 alternative file formats that are presently supported: POD5 is a signal data file format used by default on instruments from Oxford Nanopore Technologies (ONT); SLOW5 is an open-source file format originally developed as an alternative to ONT's previous file format, which was known as FAST5. The choice of format may have important implications for the cost, speed, and simplicity of nanopore signal data analysis, management, and storage. To inform this choice, we present a comparative evaluation of POD5 versus SLOW5. We conducted benchmarking experiments assessing file size, analysis performance, and usability on a variety of different computer architectures. Binary SLOW5 (BLOW5) showed superior performance during sequential and nonsequential (random access) file reading on most systems, manifesting in faster, cheaper basecalling and other analysis, and we could find no instance in which POD5 file reading was significantly faster than BLOW5. We demonstrate that BLOW5 file writing is highly parallelizable, thereby meeting the demands of data acquisition on ONT instruments. Our analysis also identified differences in the complexity and stability of the software libraries for SLOW5 (slow5lib) and POD5 (pod5), including a large discrepancy in the number of underlying software dependencies, which may complicate the pod5 compilation process. In summary, many of the advantages originally conceived for SLOW5 remain relevant today, despite the replacement of FAST5 with POD5 as ONT's core file format.