EPSAPG: A Pipeline Combining MMseqs2 and PSI-BLAST to Quickly Generate Extensive Protein Sequence Alignment Profiles

Abstract

Numerous machine learning (ML) models employed in protein function and structure prediction depend on evolutionary information, which is captured through multiple-sequence alignments (MSA) or position-specific scoring matrices (PSSM) as generated by PSI-BLAST. Consequently, these predictive methods are burdened by substantial computational demands and prolonged computing time requirements. The principal challenge stems from the necessity imposed on the PSI-BLAST software to load large sequence databases sequentially in batches and then search for sequence alignments akin to a given query sequence. In the case of batch queries, the runtime scales even linearly. The predicament at hand is becoming more challenging as the size of bio-sequence data repositories experiences exponential growth over time and as a consequence, this upward trend exerts a proportional strain on the runtime of PSI-BLAST. To address this issue, an eminent resolution lies in leveraging the MMseqs2 method, capable of expediting the search process by a magnitude of 100. However, MMseqs2 cannot be directly employed to generate the final output in the desired format of PSI-BLAST alignments and PSSM profiles. In this research work, I developed a comprehensive pipeline that synergistically integrates both MMseqs2 and PSI-BLAST, resulting in the creation of a robust, optimized, and highly efficient hybrid alignment pipeline. Notably, the hybrid tool exhibits a significant speed improvement, surpassing the runtime performance of PSI-BLAST in generating sequence alignment profiles by a factor of two orders of magnitude. It is implemented in C++ and is freely available under the MIT license at https://github.com/issararab/EPSAPG.