UCSC Liftover from hg19 to chimp changes segment lengths

Question

[sskvera]

Hi Everyone,

I have a list of 1kb windows in hg19 coordinates for which I need to get the chimp sequence. Here is what I did:

1. Downloaded Chimp fasta file from UCSC
2. Lifted over bedfiles with the positions of the windows from Hg19 to PanTro4 using UCSC's liftOver tool
3. Used bedtools to extract segments from PanTro4 fasta file

There were two problems:

1. Some of the windows could not be mapped to PanTro4 (~7%)
2. Many of the windows changed size during the liftOver from Hg19 to PanTro4 (~60%)

These windows are already filtered for problematic sites (hypermutable sites and sites of uncertain orthology with chimpanzee), so I did not expect such a high ratio of windows to be problematic during the liftover step. Does anyone have an idea of what is happening here?

Many thanks in advance for any advice!

[WouterDeCoster]

Perhaps I misunderstood, but you used both hg18 and hg19?

[GenoMax]

@sskvera: Did you forget to include hg18 --> hg19 liftover in the description above (which you must have done)?

[sskvera]

Sorry, should've been hg19 (corrected post).

Answer 1

[microfuge]

I don't think it is recommended to use liftover to convert between species. UCSC table browser could extract chip regions from the bed file from the 'Comparative Genomics' tracks using table browser intersect. An alternate way would be to download whole genome alignment from UCSC and extract the chimp sequence/coordinates.

[GenoMax]

That is not completely true. Liftover is recommended for conversion between related species. Web interface to liftover tool shows combinations that can be used with liftover. Human <--> Chimp conversion is an available option on web (on the command line this may need to be done sequentially, depending on what chain files are available).

Should one use such a conversion is something one would need to decide based on the intended application.

[microfuge]

I usually use a command line version downloaded from UCSC and it gives this warning when invoked without arguments. Sorry did not know about the webinterface . I agree with the 'intended application' part.

---

WARNING: liftOver was only designed to work between different assemblies of the same organism. It may not do what you want if you are lifting between different organisms. If there has been a rearrangement in one of the species, the size of the region being mapped may change dramatically after mapping.

---

[sskvera]

Thank you for the replies! I am aware that there may be problems with conversions between species (although I thought that liftOver should be OK for such closely related ones, as genomax2 and microfuge mentioned). However, I am still very surprised that such a high ratio of windows were problematic (~60%!). It does not seem realistic to me that there is an indel in more than half of the 1000kb windows between humans and chimpanzees! What do you think?