Format conversion

<!-- metadata: category: bioinformatics-sequence oversight: low version: 1.0.0 author: Pradyumna Jayaram tags: [bioinformatics-sequence, research] difficulty: intermediate -->

Sequence Format Conversion

Hard rules

1. No fabricated citations. Every cited work must resolve to a verifiable
2. No claim without provenance. Every quantitative or factual claim
3. No silent failure. Every script invocation, API call, or tool use must declare its exit status and what to do on non-zero. A skill that silently swallows errors is a violation.

When to use

• Switching between FASTA and GenBank for a tool that requires one or the other.
• Converting a directory of .gb → .fasta for BLAST.
• Converting EMBL → GenBank for NCBI submission compatibility.
• Stripping quality from FASTQ for tools that need FASTA.

When NOT to use

• Format conversions that need re-validation (e.g., a GenBank you got from a third party — see the bio-format-validation skill from read-qc).
• Lossy round-trips where you actually need the annotation back: don't go FASTA → GenBank and expect features.

Prerequisites

• biopython>=1.83
• For large-scale batch: seqkit (C) is far faster than Python.

Core workflow

1. Decide if the conversion is lossless. Cross-format only when features don't matter, or when both formats are annotation-aware.
2. Use SeqIO.convert(src, src_fmt, dst, dst_fmt) — it returns the record count and handles buffered I/O.
3. For streaming, prefer parse + write when you need to filter or modify on the way through.
4. Always set explicit formats — never rely on extension sniffing alone.

Lossless vs. lossy

From → To	Sequence	Annotation	Quality
FASTA → GenBank	✓	✗ (none to start)	✗
GenBank → FASTA	✓	✗ (dropped)	✗
GenBank → EMBL	✓	✓	✗
EMBL → GenBank	✓	✓ (with caveats)	✗
FASTQ → FASTA	✓	n/a	✗
FASTQ → FASTQ (re-encode)	✓	n/a	✓ (re-encoded to Phred+33)
GFF3 → BED	n/a	partial (gene-level only)	n/a

Code patterns

GenBank → FASTA (annotation dropped)

from Bio import SeqIO

n = SeqIO.convert("input.gb", "genbank", "out.fasta", "fasta")
print(f"Converted {n} records (annotation discarded)")

FASTA → GenBank (skeleton record, no features)

from Bio import SeqIO
from Bio.SeqRecord import SeqRecord
from Bio.Seq import Seq

recs = []
for r in SeqIO.parse("input.fasta", "fasta"):
    recs.append(SeqRecord(Seq(str(r.seq)), id=r.id, description=r.description,
                          annotations={"molecule_type": "DNA"}))

SeqIO.write(recs, "out.gb", "genbank")

Note: NCBI's tbl2asn will reject records without molecule_type and other minimal metadata. Add a translation table annotation if you'll submit.

EMBL → GenBank

from Bio import SeqIO

n = SeqIO.convert("input.embl", "embl", "out.gb", "genbank")
print(f"Converted {n} records (qualifiers mapped to GenBank qualifiers)")

EMBL → GenBank is lossy in edge cases (e.g., /translation table differences); always re-validate before submission.

FASTQ → FASTA (drop quality)

from Bio import SeqIO
SeqIO.write(SeqIO.parse("reads.fastq", "fastq"), "reads.fasta", "fasta")

FASTA → FASTQ (dummy Q-scores)

There's no canonical way — FASTQ requires quality. Common convention: write a constant Q40 (ASCII I):

from Bio import SeqIO
from Bio.SeqRecord import SeqRecord
from Bio.Seq import Seq

def fasta_to_fastq_q40(in_fa: str, out_fq: str):
    with open(out_fq, "w") as out:
        for r in SeqIO.parse(in_fa, "fasta"):
            quals = [40] * len(r.seq)
            r.letter_annotations["phred_quality"] = quals
            SeqIO.write(r, out, "fastq")

Batch convert a directory

from pathlib import Path
from Bio import SeqIO

for gb in sorted(Path("genbank").glob("*.gb")):
    out = Path("fasta") / (gb.stem + ".fasta")
    n = SeqIO.convert(str(gb), "genbank", str(out), "fasta")
    print(f"{gb.name} -> {out.name}: {n}")

Streaming with a filter (length ≥ 500 bp)

from Bio import SeqIO

def long_records(path, fmt, min_len=500):
    for r in SeqIO.parse(path, fmt):
        if len(r.seq) >= min_len:
            yield r

SeqIO.write(long_records("input.gb", "genbank"), "long_only.fasta", "fasta")

Re-encode legacy FASTQ to Phred+33

from Bio import SeqIO

records = SeqIO.parse("legacy.fastq", "fastq-illumina")
SeqIO.write(records, "modern.fastq", "fastq")

Common pitfalls

• GenBank round-trip loses qualifiers not mapped 1-to-1. Always diff before assuming a clean round-trip.
• SeqIO.convert returns 0 silently if the source format is wrong. Always assert n > 0 in pipelines.
• FASTQ → FASTA can produce files with non-IUPAC characters if the FASTQ had ambiguity codes. Decide whether to clean first.
• molecule_type annotation missing → GenBank writers will warn or reject.
• Don't write features onto a Seq from a SeqIO.parse(... "fasta") record — the seq is a Seq but the record's features list is empty by design.

Validation

• After conversion, grep -c '^>' out.fasta equals the record count.
• For GenBank → FASTA → GenBank, compare Bio.SeqIO.parse() record counts and IDs.
• For FASTQ re-encoding, re-parse with fastq and confirm letter_annotations["phred_quality"] are all 40 in the dummy case.

Open alternatives

Need	Tool
Bulk FASTA → GenBank for tbl2asn	NCBI tbl2asn
Sequence-only format conversion at scale	seqkit convert (C, very fast)
FASTA quality inspection	seqkit fx2tab
Annotation-aware conversion	gff3_to_genbank from pycbio

References

• Biopython convert formats table: https://biopython.org/wiki/SeqIO#File_Formats
• INSDC feature table: https://www.insdc.org/submitting-standards/feature-table/
• Companion: ors-bioinformatics-sequence-batch-sequence-processing, ors-bioinformatics-sequence-read-write-sequences.

Changelog

• 1.0.0 (2026-06-10): Initial adaptation by Pradyumna Jayaram from bio-format-conversion (bioSkills-main/sequence-io/format-conversion).

Cross-references

Other skills in this category:

• batch-processing
• bowtie2-alignment
• bwa-alignment
• bwa-mem2-alignment
• codon-usage
• compressed-sequence-files
• fastq-quality-scores
• filter-sequences
• hisat2-alignment
• motif-search
• paired-end-fastq
• pysam-genomics
• read-write-sequences
• reverse-complement
• sam-bam-basics
• samtools-bam-processing
• seq-objects
• sequence-properties
• sequence-slicing
• sequence-statistics
• star-alignment
• transcription-translation