Iso-Seq Collapse

After transcript sequences are mapped to a reference genome, isoseq collapse can be used to collapse redundant transcripts (based on exonic structures) into unique isoforms. Output consists of unique isoforms in GFF format and secondary files containing information about the number of reads supporting each unique isoform.

Collapse Examples

Execution

Map reads using pbmm2 before collapsing

pbmm2 align --preset ISOSEQ --sort <input.bam> <ref.fa> <mapped.bam>

Collapse mapped reads into unique isoforms using isoseq collapse.

For single-cell Iso-Seq:

isoseq collapse <mapped.bam> <collapse.gff>

For bulk Iso-Seq:

isoseq collapse --do-not-collapse-extra-5exons <mapped.bam> <flnc.bam> <collapsed.gff>

Notes:

• The optional <flnc.bam> input is required to get the correct FLNC counts for bulk Iso-Seq in the flnc_count.txt supplemental file.
• collapse by default will collapse isoforms containing 5p degradation as of version 3.8.0. To turn this off --do-not-collapse-extra-5exons should be used. This option is recommended for bulk Iso-Seq.

Output

• collapse.gff contains the collapsed isoforms in gff format.
• *.abundance.txt contains information about the number of FLNC reads supporting each isoform and cell barcodes if applicable. Each unique isoform has the ID format PB.X.Y, while count_fl denotes the number of unique molecules (after UMI deduplication) supporting the isoform, and fl_assoc denotes the number of reads (before UMI deduplication) supporting it. cell_barcodes shows the list of single cell barcodes from which the reads came from, if applicable. This file should be used for downstream pigeon steps for Single-cell Iso-Seq.

    pbid	count_fl	fl_assoc	cell_barcodes
    PB.1.1	2	2	ATCCATTCACCTCTGT,ATCGGCGCAGAGATGC
    PB.2.1	1	1	CGGACACCATTGCCGG
    PB.3.1	1	1	ACTTCGCGTCTAACTG
  

• *.flnc_count.txt contains information about the number of FLNC reads supporting each isoform before any clustering or deduplication. Each unique isoform has the ID format PB.X.Y and the FLNC counts will be separated by sample if multiple samples are present. This file should be used for downstream pigeon steps for Bulk Iso-Seq.

    id	BioSample1	BioSample2
    PB.1.1	2	2
    PB.2.1	1	2
    PB.3.1	1	1
  

• *.group.txt shows the grouping of redundant isoforms (based on mapped exonic structures), where the read names molecule/<number> denote a unique molecule after UMI deduplication and the read names transcript/<number> denote a clustered transcript.

    PB.1.1	molecule/7343975,molecule/7738347
    PB.2.1	molecule/14601188
    PB.3.1	molecule/3998518
  

• *.read_stat.txt shows the assignment of each read (before UMI deduplication or clustering) to the final, unique isoforms PB.X.Y. Read names with the format <movie>/<zmw>/ccs indicate a CCS read, whereas <movie>/<zmw>/ccs/<start>_<end> further denotes a segment of a CCS read (S-read), likely as a result of segmentation (using, for example, Skera) of concatenated single cell libraries.

    id	pbid
    m64012_220421_000242/120719489/ccs/10460_11196	PB.1.1
    m64012_220421_000242/17565024/ccs/13918_14203	PB.1.1
    m64012_220421_000242/161089449/ccs/1955_2888	PB.2.1
    m64012_220421_000242/158664505/ccs/2488_2901	PB.3.1
  

Collapse FAQ

As of isoseq3 v3.8.0 collapse has algorithmic updates. These updates include performance improvements and updates to isoform collapse logic.

What is new in v3.8.0 and later?

Collapsing extra 5p exons

For applications like single-cell Iso-Seq where there is a higher percentage of 5p truncated isoforms, it is useful to collapse isoforms that have a matching exon structure with the exception of extra 5p exons. Previous versions of collapse did not merge isoforms with extra 5p exons. As of v3.8.0, collapse will merge these isoforms by default. To not allow merging isoforms with extra 5p exons, use --do-not-collapse-extra-5exons. This option is used in the bulk Iso-Seq workflow.

Flexible first/last exon differences

Previous versions of collapse used stringent maximum differences (5bp) for both internal junctions and external junctions. As of v3.8.0, the maximum 5p and 3p differences have been increased and paramaters added to allow adjustments. Note: the maximum 5p difference only applies when --do-not-collapse-extra-5exons is set.

Latest v4.0.0 collapse maximum junction difference parameters:

  --max-fuzzy-junction            INT    Ignore mismatches or indels shorter than or equal to N. [5]
  --max-5p-diff                   INT    Maximum allowed 5' difference if on same exon. [50]
  --max-3p-diff                   INT    Maximum allowed 3' difference if on same exon. [100]

What if I want to use the legacy collapse logic?

The legacy collapse logic can be recreated using the following parameters:

isoseq collapse --do-not-collapse-extra-5exons --max-5p-diff 5 --max-3p-diff 5 <mapped.bam> <collapsed.gff>