This is a great 1-page explainer - perhaps a little oversimplified compared to the video, but it definitely helped me wrap my head around the process…

Enriching MSAs through SRA-scale searches improved predictions significantly [1]. Therefore, we queried all viral proteins with low MSA coverage against Logan [2]; improving the prediction of over ~10% to a pLDDT > 70.
[1] cshperspectives.cshlp.org/content/earl...
[2] www.biorxiv.org/content/10.1...