Subject: Inquiry about Using tensorQTL for SV eQTL Mapping

[C-YONG]

I am currently working on a project involving structural variants (SVs) and their association with gene expression, and I am interested in using tensorQTL for eQTL mapping involving SVs.

I would like to clarify whether tensorQTL supports the calculation of eQTLs for structural variants, or if any modifications to the script or methodology would be required to adapt it for this use case. Specifically, I am interested in:

1. Whether the software can process SVs as part of the input data.
2. Any necessary adjustments or pre-processing steps that may be needed to make SV data compatible with tensorQTL.
3. How tensorQTL handles input files for different types of genomic variants (SVs vs SNPs) and whether any modifications to the standard VCF format or other files are necessary.
   Any guidance or suggestions would be greatly appreciated.

[francois-a]

Hi, tensorQTL provides a general linear regression framework, and if you encode the SV genotypes as dosages it will work fine.

[C-YONG]

Thank you for developing such a great software! I have a quick question: I noticed that SNPs are typically considered eQTL candidates within a 1MB region. When using SVs, do we only consider the breakpoints' positions, or do we also need to take into account the start and end positions of insertions and deletions? For example, if the start and end positions of insertions or deletions overlap with a gene’s 1MB region, should they also be considered as potential eQTLs? Do I need to modify the software manually to account for this?

Thank you again for your help!

[francois-a]

This is flexible and you can define the windows as you see fit. The cis-window is defined as [start-window, end+window] — for TSS ± 1Mb typically used for eQTLs, start and end in the input BED file must be the same, but this is flexible and you could set window to 0 and define start/end for each phenotype.