LiftOver alignments are used to map annotations from one human assembly to another one.
The subtracks of this track were created by the T2T consortium using the minimap2 aligner and strong filters;
it maps CHM13 coordinates to the human assemblies hg19 and hg38.
The T2T pipeline used the minimap2
aligner which outputs long alignments that do not require "chaining" of
alignments into longer ones, then removed alignments that go to other
chromosomes and removed all alignments to alternate haplotypes, fixes
(corrections to the assembly) and unplaced contig sequences.
This means that the T2T alignments are tuned for high specificity. These alignments are
probably best used for mapping annotations to hg38 in automated pipelines and
in cases where the final processing on hg38 does not use alts/fixes/unplaced
sequences and when one wants to be sure that annotations that are mapped are
as reliable as possible.
Here is an example to illustrate the liftOver track:
- Example 1: The acrocentric
arms of chromosomes 13, 14, 15, 21, 22 and Y where not sequenced in hg38 at all but they are
present in CHM13. The T2T liftOver shows that little is mappable there, as
the sequence is entirely new.
A note on genes: As the link above shows, even though the sequence is new,
the T2T group mapped hg38 Gencode 35 gene models into these regions using
CAT/LiftOff. This is because CAT and liftOff are using approaches for their
lifting of genes / mapping that are not based on the liftOver alignments but sequence homology.
Also, we created dot plots from these alignments:
The track displays boxes joined together by either single or double lines,
with the boxes represent aligning regions, single lines indicating gaps that
are largely due to a deletion in the CHM13 v2.0 assembly or an insertion in
the GRCh38 or GRCh37, and double lines representing more complex gaps that
involve substantial sequence in both assembly.
LiftOver chain file downloads
One-to-one liftOver chain files to and from GRCh38/hg38 and GRCh37/hg19 are available here:
The mask file for GRCh38/hg38
T2T GRCh38/hg38 pre-processing
To prevent ambiguous alignments, all false duplications, as determined by the Genome in a Bottle Consortium
(GCA_000001405.15_GRCh38_GRC_exclusions_T2Tv2.bed), as well as the GRCh38 modeled centromeres,
were masked from the GRCh38/hg38 primary assembly. In addition, unlocalized and unplaced (random) contigs were removed.
T2T GRCh37/hg19 pre-processing
Unlocalized and unplaced (random) contigs were removed from the GRCh37/hg19 assembly.
T2T Alignment and Chain Creation
For the minimap2-based pipeline, the initial chain file was generated using
nf-LO v1.5.1 with
minimap2 v2.24 alignments. These chains were then split at all locations that contained unaligned segments greater than 1kbp or gaps greater than 10kbp. Split chain files were then converted to PAF format with extended CIGAR strings using chaintools (https://doi.org/10.5281/zenodo.6342391, v0.1), and alignments between nonhomologous chromosomes were removed. The trim-paf operation of rustybam (https://zenodo.org/record/6342176, v0.1.29) was next used to remove overlapping alignments in the query sequence, and then the target sequence, to create 1:1 alignments. PAF alignments were converted back to the chain format with paf2chain commit f68eeca, and finally, chaintools was used to generate the inverted chain file.
Full commands with parameters used were:
nextflow run main.nf --source GRCh38.fa --target chm13v2.0.fasta --outdir dir -profile local --aligner minimap2
python chaintools/src/split.py -c input.chain -o input-split.chain
python chaintools/src/to_paf.py -c input-split.chain -t target.fa -q query.fa -o input-split.paf
awk '$1==$6' input-split.paf | rb break-paf --max-size 10000 | rb trim-paf -r | rb invert | rb trim-paf -r | rb invert > out.paf
paf2chain -i out.paf > out.chain
python chaintools/src/invert.py -c out.chain -o out_inverted.chain
The above process does not add chain ids or scores. The UCSC utilities
chainScore are used to update the
chainMergeSort out.chain | chainScore stdin chm13v2.0.2bit hg38.2bit chm13v2.0-hg38.chain
chainMergeSort out_inverted.chain | chainScore stdin hg38.2bit chm13v2.0.2bit hg38-chm13v2.0.chain
uses dynamic programming and the CIGAR string to find an optimal
splitting point between overlapping alignments in the query sequence. It
starts its trimming with the largest overlap and then recursively trims
Results were validated by using chaintools to confirm that there were no
overlapping sequences with respect to both CHM13v2.0 and GRCh38 in the
released chain file. In addition, trimmed alignments were visually inspected
with SafFire to confirm their quality.
The T2T v1_nflo liftOver chains were generated by Nae-Chyun
Chen<email@example.com> and Mitchell
Vollger<firstname.lastname@example.org>. The UCSC liftOver chains and the dot-plots
were created by Hiram Clawson.
lastz was developed by Robert Harris, Pennsylvania State University.
The axtChain program was developed at the University of California at
Santa Cruz by Jim Kent with advice from Webb Miller and David Haussler.
The browser display and database storage of the chains and nets were created
by Robert Baertsch and Jim Kent.
The chainNet, netSyntenic, and netClass programs
were developed at the University of California
Santa Cruz by Jim Kent.
Nurk S, Koren S, Rhie A, Rautiainen M, et al. The complete sequence of a human genome. bioRxiv, 2021.
(2007) Improved pairwise alignment of genomic DNA
Ph.D. Thesis, The Pennsylvania State University
Chiaromonte F, Yap VB, Miller W.
Scoring pairwise genomic sequence alignments.
Pac Symp Biocomput. 2002:115-26.
Kent WJ, Baertsch R, Hinrichs A, Miller W, Haussler D.
duplication, deletion, and rearrangement in the mouse and human genomes.
Proc Natl Acad Sci U S A. 2003 Sep 30;100(20):11484-9.
PMID: 14500911; PMC: PMC208784
Schwartz S, Kent WJ, Smit A, Zhang Z, Baertsch R, Hardison RC,
Haussler D, Miller W.
Human-mouse alignments with BLASTZ.
Genome Res. 2003 Jan;13(1):103-7.
PMID: 12529312; PMC: PMC430961