4. Protein evidence (utils)

Both protein inputs to reconcile come from helixforge utils. You do not run any external protein tool by hand, and they share a single source FASTA:

1. utils fetch-db downloads and DIAMOND-formats the SwissProt-plants protein set.
2. utils align runs miniprot against the genome to produce the backstop GFF (--miniprot).
3. reconcile consumes the same FASTA directly as --protein-db.

One FASTA, not two protein sources

--miniprot (backstop CDS) and --protein-db (DIAMOND homology, gates Mikado) are two different reconcile flags, but here they are derived from the one swissprot_plants/uniprot_sprot_plants.fasta. There is no second protein source to obtain.

4a. Download the protein database

apptainer exec helixforge.sif helixforge utils fetch-db \
  --db swissprot_plants \
  --cache-dir .

This downloads the taxonomic-division flat file, converts it to FASTA, and DIAMOND-formats it, writing:

./swissprot_plants/uniprot_sprot_plants.fasta        # ← --proteins and --protein-db
./swissprot_plants/uniprot_sprot_plants.dmnd         # DIAMOND database

The output directory is <cache-dir>/<db-name>/, so --db swissprot_plants --cache-dir . lands the FASTA at exactly swissprot_plants/uniprot_sprot_plants.fasta, the path the align and reconcile steps expect. Confirm the available names with helixforge utils fetch-db --list (see the fetch-db reference):

--db	Contents	Output FASTA
swissprot	Curated, all organisms	swissprot/uniprot_sprot.fasta
swissprot_plants	Curated plant proteins (Viridiplantae)	swissprot_plants/uniprot_sprot_plants.fasta

Reconciling the recorded command

The run log shows fetch-db --db swissprot --cache-dir /tmp/. Two adjustments make the paths line up with what align/reconcile actually read:

• --db swissprot_plants, not swissprot. --db swissprot resolves to uniprot_sprot.fasta (all organisms) under a swissprot/ directory, not the swissprot_plants/uniprot_sprot_plants.fasta the downstream steps consume. The plant subset is swissprot_plants.
• --cache-dir ., not /tmp/. The --miniprot and --protein-db paths are relative (swissprot_plants/…). Caching into the working directory makes them resolve; with --cache-dir /tmp/ you would instead point --proteins/--protein-db at /tmp/swissprot_plants/uniprot_sprot_plants.fasta.

4b. Align proteins to the genome (miniprot)

utils align wraps miniprot to produce the protein-to-genome GFF3 that becomes reconcile --miniprot:

apptainer exec helixforge.sif helixforge utils align \
  --genome genome/Zm-B73-REFERENCE-NAM-5.0.fa \
  --proteins swissprot_plants/uniprot_sprot_plants.fasta \
  --out Zm-B73-sw-plants-miniprot.gff3 \
  --threads ${SLURM_CPUS_ON_NODE}

Flag	Value
--genome	genome/Zm-B73-REFERENCE-NAM-5.0.fa
--proteins	swissprot_plants/uniprot_sprot_plants.fasta (from 4a)
--out	Zm-B73-sw-plants-miniprot.gff3 → reconcile --miniprot
--threads	${SLURM_CPUS_ON_NODE} (the allocation's cores)

See the utils align reference for the full option list.

What feeds reconcile

1. swissprot_plants/uniprot_sprot_plants.fasta → reconcile --protein-db (DIAMOND homology; gates Mikado + isoform discovery).
2. Zm-B73-sw-plants-miniprot.gff3 → reconcile --miniprot (backstop CDS only).

Next: run the Helixer gene prediction →