cdskit filter

cdskit filter removes CDS records that fail sequence-level quality rules.

Example

cdskit filter --seq_file input.fasta --out_file filtered.fasta --drop_internal_stop yes --min_clean_codon_fraction 0.8 --dedup keep-longest

What can be filtered

• Non-triplet sequence lengths.
• Sequences with internal in-frame stop codons.
• Sequences whose clean-codon fraction is below a threshold.
• Duplicate IDs, resolved by keep-first or keep-longest.

Key options

• --drop_non_triplet yes|no (default yes): Drop sequences whose length is not divisible by three.
• --drop_internal_stop yes|no (default no): Drop sequences with internal stop codons.
• --min_clean_codon_fraction FLOAT (0–1, default 0.5): Minimum fraction of clean codons required to retain a sequence.
• --dedup no|keep-first|keep-longest (default no): Duplicate-ID handling after filtering.
• --report PATH: Optional JSON or TSV report listing kept and dropped IDs.

A clean codon contains no missing character (-, ?, or .), no ambiguous nucleotide (N, R, Y, etc.), and no definite stop codon under ordinary translation. A dual-coding A/C/G/T codon is clean but is also reported as context-dependent.

TSV report format

TSV reports use schema version 2. Every row starts with schema_version and section. Use summary for run settings/counts, id_set for kept or dropped ID lists, drop_reason for per-reason counts, and sequence for one row per input record. List-valued cells such as ids and drop_reasons contain compact JSON arrays.

Notes

• Input sequences must be DNA.
• The command filters whole sequences, not alignment columns.

Codon semantics

See codon meaning and uncertainty for definite versus possible stops, dual-coding tables 27/28/31, overlapping ambiguity/stop counts, and report compatibility. These rules describe sequence QC, not gene function.

Citation

No separate method paper is designated for this utility. Report the CDSKIT version, command and options; see citing CDSKIT.