Add matched WT to pVACfuse predictions

In pVACseq we require at least 4 overlapping amino acids between a matched WT and MT epitope. For inframe fusions in pVACfuse we could apply a similar approach by matching epitopes on the left side of the fusion to the 5" transcript and epitopes to the right of the fusion to the 3" transcript. For frameshift fusion we can match epitopes on the left side of the fusion to the 5" transcript, similar to frameshift variants in pVACseq. We would require at least 4 overlapping AAs on one end or the other; otherwise there wouldn’t be a matched WT.

This would require us to look up WT transcripts of the 5" and 3" transcripts at runtime. I believe Ensembl has an API for that (see https://rest.ensembl.org/documentation/info/sequence_id and https://ensemblrest.readthedocs.io/en/latest/#). We could probably also use the peptide fasta, if provided for the reference proteome step, and use the API as a fallback.

Start by tracing the pVACfuse prediction flow and comparing it with pVACseq’s matched-WT handling. Evaluate runtime lookup of the 5′ and 3′ WT transcripts through the Ensembl sequence API, with the peptide FASTA as a possible fallback. Done means defining and implementing the stated overlap rule for inframe and frameshift fusions.