How does the epitope homology search work?

When searching for peptidic epitopes by amino acid sequence, the IEDB offers the option to retrieve epitopes homologous to your input sequence at 70%, 80% or 90% identity, rather than only exact or substring matches. This is done in two steps. First, BLAST is used to broadly retrieve any candidate epitopes similar to the input. Secondly, the selected identity threshold is applied as a post-processing filter, calculated relative to the shorter one of the two epitopes being compared. The identity cut-off you see in the interface is therefore applied after the BLAST step, not by BLAST itself.

The search uses blastp from NCBI BLAST+ 2.13.0, run against a FASTA database of the epitope sequences in the IEDB, with the following call:
blastp -query <input> -db <iedb database> -evalue 100 -ungapped -seg yes -matrix PAM30 -comp_based_stats F -num_descriptions 1000000 -num_alignments 1000000

For more information on searching epitopes by sequence, see IEDB Epitopes.