Inconsistency between UCSC liftover webtool and pyliftover module

Question

[younglin113]

Hello everyone,

I'm trying to convert the genome coordinates from mm10 to hg38, since I run into some problems installing liftOver in my Linux server, so I instead used pyliftover for this. To test this tool, I input 9 intervals in total, 4 of them cannot be properly converted.

To confirm the results, I also uploaded the same 9 intervals to the online UCSC liftOver, while only 1 of them cannot be converted.

I'm wondering why don't they have the same results? Has anyone ever noticed this difference? I appreciate your help.

Answer 1

[Randy H]

Maybe you are using terminology that does not express what you mean? If not, then pyLiftover works on single, point coordinate values. UCSC Liftover works on intervals. You cannot convert the endpoints of an interval in pyLiftover and expect to get the same result as UCSC when supplying it with the interval itself. See the pyLiftover README.md. As it is, UCSC suggests not doing a point liftover. Instead, you should use the coordinates for each build that are stored with an rsID. UCSC otherwise only takes intervals (although an interval of length 1 is allowed in the tool UI).