Hello, I am trying to build a custom SnpEff database for Saccharomyces cerevisiae using the latest SGD release (R64-4-1_20230830).

My input files are:

GFF: saccharomyces_cerevisiae_R64-4-1_20230830.gff.gz

Genome FASTA: S288C_reference_sequence_R64-4-1_20230830.fsa.gz

Protein FASTA: orf_trans_all_R64-4-1_20230830.fasta.gz

CDS FASTA: orf_coding_all_R64-4-1_20230830.fasta.gz

The Problem: When running java -jar snpEff.jar build -gff3 -v cerevisiae_ownversion, my Protein check divides exactly into: OK: 2067 Not found: 4512 (Totaling 6,579 coding transcripts).

The IDs match up for exactly 2,067 genes, but 4,512 are dropping to "Not found" due to header/attribute mismatch between the GFF structural fields and the FASTA sequence names. If I try to filter out the non-coding structural parent attributes or drop lines to force a clean 0 "Not found" metric, SnpEff throws a FATAL ERROR: Most Exons do not have sequence or fails to map them completely.

Has anyone successfully built a custom database for this specific SGD release without getting thousands of "Not found" proteins? Is there a standard regex or configuration tweak needed to get SnpEff to parse the _mRNA and _id001 identifiers consistently?