Quick start

Prerequisites

Install the package and external get_orfs executable as described in Installation. The first encoder invocation downloads a model unless it is already cached.

Complete pipeline

genome_entropy run --input genome.fasta --output results.json

The default gbouras13/modernprost-50M model predicts 3Di and 12-state encodings. The JSON contains raw entropy and uses unified schema 2.1.0.

GenBank input enables CDS matching:

genome_entropy run --genbank genome.gbk.gz --output results.json

Supplying both --input and --genbank uses sequences from FASTA and CDS annotations from GenBank. IDs must correspond for useful matching.

Step-by-step workflow

genome_entropy orf --input genome.fasta --output orfs.json
genome_entropy translate --input orfs.json --output proteins.json
genome_entropy encode3di --input proteins.json --output structures.json
genome_entropy entropy --input structures.json --output entropy.json

The standalone entropy report cannot reconstruct whole-contig entropy and writes dna_entropy_global as 0.0. Prefer run when the unified complete record is required.

Protein-only workflow

genome_entropy fasta-to-protein --input proteins.faa --output proteins.json
genome_entropy encode3di --input proteins.faa --output structures.json

encode3di accepts protein FASTA directly, so conversion is only needed when an intermediate protein-record JSON is useful.

Model and device selection

genome_entropy run --input genome.fasta --output results.json \
    --model gbouras13/modernprost-base --device cuda
genome_entropy encode3di --input proteins.faa --output structures.json \
    --multi-gpu --gpu-ids 0,1 --encoding-size 8000

Use the 50M default for ordinary examples; the larger model needs more memory. Estimate a safe token budget on the target hardware with estimate-tokens.

Next steps

Read Models and structural-state encodings for encoder details, Data formats, coordinates, and entropy for schemas and normalised entropy, Command-line reference for every option, and Machine-learning workflow before training or interpreting a classifier.