Hello Kian, It is a known that liftOver is a great tool for the majority of mappings, but also that occasionally a human can detect that there is a better map location.
To determine the root cause of your mapping issues, these influencing factors could be investigated: 1 - size of flanking region used to place the SNP 2 - presence of repeat sequence in genome or flanking sequence 3 - legitimate or artifact genome duplications 4 - unresolved assembly issues in either of the reference genomes Have you tried using BLAT to see if the results of liftOver agree? Some users find this helpful. The parameters for liftOver can also be tuned to capture more map locations, so would be a choice of mappings to pick from. Perhaps take the 1% with initial mappings that you think are incorrect and do a second liftOver run, lowering the match identity and allowing for multiple mappings. Adding in a minimum chain length is also recommended for this type of run, to help reduce the noise. Conservative min chain length would be ~500, but ~1000 is also a good choice that is unlikely to remove useful mappings. If you want to send a specific example or two, we can provide feedback. Include all of the details - the data, the liftOver tool type (web or line command), all exact parameters, and what you think is wrong with the output, so that we can duplicate what you are doing. Hopefully this provides some useful help/advice, Thanks, Jen UCSC Genome Browser Suppport On 7/28/10 10:06 AM, Kian Preston-Suni wrote: > Hello, > > I have been running into problems with the liftover tool output. I am > submitting SNPs from WGS to liftover to convert from hg19 to hg18 positions, > and around 1% map back incorrectly to the hg 18 position. For example, on > chromosome 6 a large number of SNPs are mapped back to chromosome 3, 15 and > to chr6_random. Has anyone ever noticed a problem like this before? Any > insight into what could be happening here would be greatly appreciated. > > Thanks, > > Kian Preston-Suni > Dept. of Internal Medicine and Molecular Genetics > University of Michigan Medical School > _______________________________________________ > Genome maillist - [email protected] > https://lists.soe.ucsc.edu/mailman/listinfo/genome _______________________________________________ Genome maillist - [email protected] https://lists.soe.ucsc.edu/mailman/listinfo/genome
