How this file was created? All proteins (n=66,585,678) were retrieved from IMG/VR v3 database (https://genome.jgi.doe.gov/portal/IMG_VR/IMG_VR.home.html) version IMG_VR_2020-10-12_5.1 (https://doi.org/10.1093/nar/gkaa946). We used MMseqs2 (https://doi.org/10.1038/nbt.3988) for similarity-based clustering with a threshold of 70% identity (using default greedy mode and 80% reciprocal coverage of target and query). We then extracted one representative sequence per cluster (n=16,555,061) to build the IMGVR70.fa file.