RelocaTE3 identifies transposable element (TE) insertion polymorphisms from short-read resequencing data, at single-base resolution, by comparison to a reference genome. It is a modern, pure-Python reimplementation of RelocaTE2 that depends only on minimap2 and samtools (plus pysam/biopython) — no blat, bwa, bowtie2, seqtk, or Perl.

Status: reference implementation. The full pipeline (read trimming → genome re-alignment → non-reference insertion calling → reference/shared insertions → genotyping) runs end to end. On the rice Chr3 2 Mb benchmark it recovers ~178/200 simulated insertions (~89% recall, ~90% precision); see tests/acceptance_test.py. See PLAN.md for the roadmap (Rust acceleration and Nextflow scatter are planned).

With pixi (recommended — pins minimap2/samtools/bedtools):

pixi install
pixi run relocaTE3 --help

Or into an existing environment that already provides minimap2 and samtools:

pip install -e .

Run the whole pipeline for one paired-end sample:

relocaTE3 run \
  --r1 reads_1.fq.gz --r2 reads_2.fq.gz \
  --te RiceTE.fa \
  --genome reference.fa \
  --outdir HEG4_out --sample HEG4 \
  --threads 8 --mismatch 2 \
  --repeatmasker reference.fa.RepeatMasker.out \
  --genotype

Each step is also a subcommand, so a workflow engine can scatter them:

Run relocaTE3 <command> --help for options.

• Reads: one (single-end) or two (paired-end) FASTQ files (.fq/.fq.gz).
• TE library (--te): FASTA of TE/repeat consensus sequences.
• Genome (--genome): reference genome FASTA.
• RepeatMasker .out (--repeatmasker, optional): reference TE annotation, enabling reference/shared insertion calls and false-positive filtering.

Under --outdir:

flanking/        trimmed flanking reads (FASTQ)
te_portions/     TE-matching read portions (FASTA)
te_containing/   read → TE assignment table
genome_aln/      flanking/support/full reads aligned to the genome (BAM)
existingTE.bed   reference TE annotation (when --repeatmasker given)
results/
  <sample>.all_nonref_insert.gff / .txt          non-reference insertions
  <sample>.all_ref_insert.gff / .txt             reference/shared insertions
  <sample>.all_nonref_insert.characTErized.gff   genotyped insertions (--genotype)

The non-reference GFF carries the RelocaTE2 attribute set: TSD, Name (TE family), Note, and Left/Right_junction_reads and Left/Right_support_reads.

pixi run test          # run the test suite
pixi run pytest tests/acceptance_test.py   # the benchmark acceptance gate