Marc Stevens demonstrated MD5 collision in any byte-set in 2020 and published his tool for it. Hypothetically, someone can
now generate a collision in the FASTA AA charaster set and try to slip the result into gtdb_proteins_aa_reps. The resultant sequence would then be masked by the MD5 collision, making one of the colliding copies "invisible".
(It's like I said, hypothetical. The act is mostly pointless and unlikely to cause any major damage. Sure someone can make a collision in the FASTA character set, but to make the colliding sequence have its own function...)
Marc Stevens demonstrated MD5 collision in any byte-set in 2020 and published his tool for it. Hypothetically, someone can
now generate a collision in the FASTA AA charaster set and try to slip the result into
gtdb_proteins_aa_reps. The resultant sequence would then be masked by the MD5 collision, making one of the colliding copies "invisible".(It's like I said, hypothetical. The act is mostly pointless and unlikely to cause any major damage. Sure someone can make a collision in the FASTA character set, but to make the colliding sequence have its own function...)