Alternatives To Liftover

Question

[woemler]

Has anybody had any success with any tools other than LiftOver or NCBI's Genome Remapping Service for mapping/translating reference genome positions? My experience with LiftOver has been less than satisfactory and NCBI does not seem to offer any local version of their tool. Other than BLASTing every sequence I have against a full-genome database, are there any accurate alternatives?

[Pierre Lindenbaum]

what's wrong with liftOver ?

[Gjain]

Can you give an example of what and why you not satisfied with UCSC LiftOver ?

[Malachi Griffith]

Relevant post that describes coordinate conversion tools: Converting genome coordinates from one genome version to another (UCSC liftOver, NCBI Remap, Ensembl API)

[woemler]

In my experience with LiftOver, I have only had about an 80% success rate with translating chromosomal positions between reference genomes. The ease of use of the command line application is great, but it just is not getting the job done.

NCBI's tool is successful more like 95% of the time, but on more than one occassion I have wound up with two bases in the same codon ending up 10 bases apart.

[Sean Davis]

If you are starting with genomic sequence, aligning to the reference genome is probably the most fool-proof way to go.

Answer 1

[Bert Overduin]

I don't know if it would make a difference, but you could try Ensembl's Assembly Converter. It has a web interface, but is also available as a stand-alone Perl script.

[woemler]

Thanks, I'll give this one a shot next time I am at work!

Answer 2

[Devon]

There is already a local version of liftOver available. It's easy enough to convert over large datasets with it in a relatively short amount of time.

[woemler]

I do like thes version of LiftOver, it is easy to setup and automate large batch runs. As I mention above though, I find that it fails to translate a significant amount of what I throw at it.

[Malachi Griffith]

Have you brought this to the attention of the folks at UCSC? I'm curious which genome builds you are dealing with. Between recent versions of the human genome I typically get well over 95% recovery. In my most recent test of liftover I lost 2 of ~3,500 SNP positions when lifting over from hg18 to hg19 for example...

Answer 3

[bw.]

I recently put up https://liftover.broadinstitute.org/ to make it a little easier to check liftover coords for specific variants or intervals. It takes fewer clicks and supports more input formats than UCSC liftover and Assembly Converter.

[Thanujay S]

Hey! Would you consider building an API for the same? Is that something you have planned?

Answer 4

[cmdcolin]

https://github.com/agshumate/Liftoff

https://github.com/wurmlab/flo

Answer 5

[Maximilian Haeussler]

You should give pslMap from the UCSC tools collection a try. It's different that it not only lifts the ends of the regions.

If you don't have enough recovery, you should lower the minMatch parameter. By default it's set to 0.95 which might be too high for non-human liftovers. If there is a good alignment, liftOver will use it. We would need an example to find out why your region did not lift.

Answer 6

[zasnuty]

GeneBe has a simple liftOver tools: https://genebe.net/tools/liftover , pretty convenient if you wish to lift a couple of entries. I like it for the wide range if input accepted. At the moment only hg38 and hg19 is supported.