Command line

The package installs a single entry point, peagent-tool. Every command accepts --species and, where relevant, --model-path / --metadata-path to override asset resolution.

peagent-tool --help

assets

Report where the model weights and cell-group metadata resolve from, and whether they are already present. Loads nothing.

peagent-tool assets --species maize

list-celltypes

The valid tissue_celltype groups — the exact strings accepted as --target elsewhere.

peagent-tool list-celltypes --species maize --out maize_celltypes.tsv

predict-celltypes

Predicted accessibility per cell type, as a long ranked table.

peagent-tool predict-celltypes \
  --species maize \
  --fasta my_sequences.fa \
  --mode exact \
  --top-k 10 \
  --out predictions.tsv

Option	Default	Meaning
--sequence / --fasta	—	Raw sequence or FASTA text, or a FASTA file (mutually exclusive)
--mode	exact	exact or fast
--top-k	all	Keep only the top k groups per sequence
--batch-size	32	Sequences per model batch
--out	stdout	TSV output path

attribution

One contribution map, written as matching PDF and PNG.

peagent-tool attribution \
  --species maize \
  --fasta my_acr.fa \
  --target crownRoot_cortex \
  --normalize tissue \
  --plot-start 423 --plot-end 923 \
  --out-prefix cortex_contribution

Option	Default	Meaning
--target	required	global or one exact tissue_celltype
--normalize	none	none for raw scores; tissue to subtract the tissue mean (browser convention)
--plot-start, --plot-end	whole window	1-based positions on the normalized 1,344 bp input
--position-offset	0	Added to axis labels only, for genome coordinates
--position-batch-size	64	Positions mutated per model batch
--title	auto	Custom plot title

--normalize tissue requires a tissue_celltype target; global has no tissue.

score-variants

REF/ALT scoring of biallelic SNVs against a genome FASTA.

peagent-tool score-variants \
  --species maize \
  --variants variants.tsv \
  --fasta Zm-B73-REFERENCE-NAM-5.0.fa.gz \
  --target all \
  --out variant_effects.tsv

The variant table needs chrom, pos (1-based), ref, alt, and optionally variant_id:

chrom   pos ref alt variant_id
chr8    176677455   T   C   chr8_176677455_T_C

--target may be repeated, and accepts all, global or an exact tissue_celltype.

score-ref-alt

The same scoring for allele pairs you constructed yourself.

peagent-tool score-ref-alt \
  --species maize \
  --ref-fasta ref.fa \
  --alt-fasta alt.fa \
  --target global \
  --out designed_effects.tsv