MSA Search NIM supports GPU-accelerated Multiple Sequence Alignment (MSA) of a query amino acid sequence against a set of protein sequence databases. These databases are searched for similar sequences to the query and then the collection of sequences are aligned to establish similar regions even when the proteins have different lengths and motifs. The outputs of the MSA process are used to inform structural prediction models such as AlphaFold2 and OpenFold. This tends to improve structural prediction accuracy because similar sequences often have similar structures. MSA Search is also used by evolutionary biologists to look for homology between protein sequences that may indicate a common evolutionary origin. The MSA NIM implements two search styles. The AlphaFold2 search type was first used in the AlphaFold2 paper in Nature and performs a single-pass search per database. The ColabFold search process in the MSA Search NIM was first introduced in ColabFold and implements a cascaded search of generated profiles, providing even higher sensitivity and generally better throughput. Both methods utilize GPU-accelerated MMSeqs2 for improved accuracy and reduced latency. Combined with AlphaFold2 or OpenFold, the MSA Search NIM enables a sensitive and high-throughput protein structure prediction pipeline.
