Two-pass mode

Two-pass alignment improves recovery of reads spanning novel splice junctions — those not present in the GTF (or when no GTF is supplied). rustar-aligner runs both passes in a single command, matching STAR’s --twopassMode Basic.

How it works

1. Pass 1 aligns all reads (or a subset, see --twopass1readsN) against the original index, with output, chimeric detection and quantification switched off. Its junctions go through the SJ.out.tab filters and are written to <prefix>_STARpass1/SJ.out.tab.
2. Every junction in that file is inserted into the genome, as STAR does: the junction flanks are added to the genome and the suffix array and SAindex are updated, so pass-2 seeds can cross the new junctions.
3. Pass 2 re-aligns all reads against the extended genome. In pass 2 the new junctions count as annotated, as in STAR (sjdbScore, column 6 of SJ.out.tab).

Basic command

rustar-aligner \
  --genomeDir /path/to/genome_index \
  --readFilesIn reads_1.fq.gz reads_2.fq.gz \
  --readFilesCommand zcat \
  --twopassMode Basic \
  --outSAMtype BAM SortedByCoordinate \
  --outFileNamePrefix sample_

Combining with a GTF

You can use two-pass mode together with --sjdbGTFfile. The GTF’s junctions are inserted before pass 1; the pass-1 junctions are then added to them. When the same junction comes from both, the genome’s and the GTF’s version is kept, as STAR does.

rustar-aligner \
  --genomeDir /path/to/genome_index \
  --readFilesIn reads_1.fq.gz reads_2.fq.gz \
  --readFilesCommand zcat \
  --sjdbGTFfile gencode.v45.gtf \
  --twopassMode Basic \
  --outSAMtype BAM SortedByCoordinate \
  --outFileNamePrefix sample_

Limiting pass 1

For very large datasets you can limit pass 1 to a subset of reads:

--twopass1readsN 1000000

This processes the first million reads in pass 1, then runs pass 2 on the full input. The default (-1) uses all reads in both passes.

Junction filter parameters

The defaults match STAR. The most relevant filters:

• --outSJfilterOverhangMin — min overhang per motif category [noncanon, GT/AG, GC/AG, AT/AC]. Default 30 12 12 12.
• --outSJfilterCountUniqueMin — min unique-mapping reads per motif. Default 3 1 1 1.
• --outSJfilterCountTotalMin — min total reads per motif. Default 3 1 1 1.
• --outSJfilterIntronMaxVsReadN — junction is filtered when intron length exceeds the threshold for its read count. Default 50000 100000 200000 (1, 2, 3+ supporting reads).

See the CLI parameters reference for the full list.

When to use it

Use two-pass mode when:

• Working with a species or sample where the GTF annotation is incomplete
• You expect novel junctions (cancer transcriptomes, non-model organisms, single-cell)
• You want to recover the maximum number of cross-junction reads

Skip it when:

• Speed matters more than novel junction sensitivity
• You have a high-quality, complete GTF and don’t need novel discovery

Output

Two-pass mode writes the same set of files as a single-pass run, plus <prefix>_STARpass1/ (pass-1 SJ.out.tab and Log.final.out) and <prefix>_STARgenome/ (the inserted junctions; see --sjdbInsertSave). The main SJ.out.tab comes from pass 2.