As other commenters point out, this "analysis" is a joke. An E()-value of 282 says we would expect a match of this short sequence against a completely random l

As other commenters point out, this "analysis" is a joke. An E()-value of 282 says we would expect a match of this short sequence against a completely random library of the same size 282 times. Even with BLASTp's parameter correction for short sequences, NO short sequence will generate a statistically significant score. Let's try a more sensible approach. Let's see how often the sequence occurs in other databases. The author's 1/20^6 (1/64,000,000) probability calculation omits the fact that every database search considers many millions of possible alignments (a crude calculation would be 6 * the database length, or 6 * 1.6E7/20^6 ~ 10000E6/64E6 = 156 times by chance), and we're shown that in a 5,000,000 protein/1.6E7 residue database, 100% matches are seen several dozens of times, as expected. How about other protein sets? Search the human protein set (19,000 proteins), there are no 6 amino acid 100% identical matches. The best matches are 100% of 5 residues, or 83% identity over 6 (with an E()-value of 22). How many proteins do we need to search to find an exact full-length match? (we have dozens of matches in a 5,000,000 viral protein database. and none in 19,000). If you search all the NCBI landmark sequences (~500,000, non-redundant), there are 3 identical matches (E()-value: ~44) in Drosophila, Corn, and Strep. pneumoniae. So this sequence is not very special -- as the E()-value statistics indicate, it happens all the time by chance. (If it happens 3 times in a 500,000 sequence non-redundant database by chance, we would expect it to happen 30-100's of times in a 5,000,000 redundant viral database by chance, which is what we see.) Six amino acid non-significant matches are evidence that properly done statistics are accurate, and pretty much nothing else.