MMseqs2 enables sensitive protein sequence searching for the analysis of massive data sets

By Martin Steinegger, Johannes Söding. 2017

2,996 citations recorded by the source corpus.

VOLUME 35 NUMBER 11 NOVEMBER 2017 NATURE BIOTECHNOLOGY performance was to combine the doublematch criterion with making k-mers as long as possible, which required finding similar and not just exact k-mers. This effectively bases our decision on up to 2 × 7 = 14 residues instead of just 2 × 3 in BLAST or 12 letters on a size-11 alphabet in DIAMOND. MMseqs2 is parallelized on three levels: time-critical parts are manually vectorized, queries can be distributed to multiple cores, and the target database can be split into chunks distributed to multiple servers. Because MMseqs2 needs no random memory access in its innermost loop, its runtime scales almost inversely with the number of cores used (Supplementary Fig. 2). MMseqs2 requires 13.4 GB plus 7 bytes per amino acid to store the database in memory, MMseqs2 enables sensitive protein sequence searching for the analysis of massive data sets

Subfields: Ecology, Molecular Biology, Computational Biology & Bioinformatics

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Ecology

1. Shitong Luo
2. Burkhard Rost
3. Michael Heinzinger
4. Christian Dallago
5. David Baker
6. Ahmed Elnaggar
7. Gianluca Pollastri
8. Quan Le

In Molecular Biology

1. Yaoqi Zhou
2. Michael Heinzinger
3. Kuldip K. Paliwal
4. Burkhard Rost
5. Jack Hanson
6. Gianluca Pollastri
7. David Baker
8. Jens Meiler

In Computational Biology & Bioinformatics

1. Michael Heinzinger
2. Burkhard Rost
3. Gianluca Pollastri
4. Jens Meiler
5. Yaoqi Zhou
6. Ivan V. Anishchenko
7. David Baker
8. J. Dauparas

Citations in the corpus

• Robust deep learning based protein sequence design using ProteinMPNN (2022, positive): “Sequences were clustered at 30% sequence identity cutoff using mmseqs2 (10) resulting in 25,361 clusters.”
• Sampling alternative conformational states of transporters and receptors with AlphaFold2 (2022, positive): “First, all MSAs were obtained using the MMSeqs2 server ( Steinegger and Söding, 2017 ), rather than the default databases.”
• SPOT-Disorder2: Improved Protein Intrinsic Disorder Prediction by Ensembled Deep Learning (2019, methodology): “MMSeqs2 [63] is considered in this study due to its speedup in generating profiles over PSI-BLAST.”
• Modeling aspects of the language of life through transfer-learning protein sequences (2019, positive): “0 when counting the time it needs to retrieve the evolutionary information summarized in alignment profiles although using the fastest available alignment method, namely MMseqs2 [36] which already can reach speed-up values of 100-times over PSI-BLAST [33].”
• ProtTrans: Towards Cracking the Language of Life’s Code Through Self-Supervised Deep Learning and High Performance Computing (2020, positive): “353 proteins with a median sequence length of 415 residues) using our protein LMs or mmseqs2 [61], the fastest tool to gather evolutionary information from protein sequence databases at the moment.”
• ProtTrans: Towards Cracking the Language of Life’s Code Through Self-Supervised Learning (2021, neutral): “7. https://github.com/agemagician/ProtTrans/tree/master/Benchmark”