This is a selection of 70M antibody sequences from the OAS database. OAS was first filtered using the criteria described by Leem et al. (Deciphering the language of antibodies using self-supervised learning.) and then 70M sequences were randomly sampled from all repertoires.
