Custom annotations in final gene diagnosis output but not in LD feather files or in Manhattan

[Prakrithi-P]

Hi again Leon,
I tried running gsMAP with custom gtf annotations. The new features are in the Gene_Diagnostic_info file (PCC is calculated) but somehow not incorporated in the generate_ldscore step and not found in the feather files and hence not in the manhattan plot. Could you hep with what step I should be looking into? I think it is something to do with the mk_score file (.feather) generating step. I did try running manually until overlapping bim_pr and gtf_pr step and my annotations do exist in that output. So not sure where it is going wrong.

Thanks,
Prakrithi

[LeonSong1995]

Hi Prakrithi,

The step for generating the mk_score file does not require the GTF file. I’m not sure I fully understand your question. Could you provide some screenshots of the output or some test code to help clarify?

Best,
Leon

[Prakrithi-P]

Hi Leon,
I mean, I have added some custom annotations of several transcripts we are studying into the GTF. The name of these transcripts are found in the final gene diagnostic file that has PCC (GSS Correlation with GWAS P-value) and also in the mkscore file. But in the generate_ldscore/SNP_gene_pair/SNP_gene_pair_chr{chrom}.feather files generated by generate_ldscore.py. I don't find any of my custom annotations, and hence they are not in the Manhattan plot data either. I believe something went wrong in the step where SNPs are overlapped with genes. I tried running until Overlaps_gtf_bim() in generate_ldscore.py, manually, and I can see that my gene IDs are present. But I don't understand why when I run quick mode my feather files are missing these annotations. This would mean that the new annotations are not accounted for in the association P-value spot-level statistic as well. But just their GSS is correlated with the computed p-value calculated. Hope my question is clear now?

[Ganten-Hornby]

Are you referring to the use of a custom GTF file? In the quick mode, we employ precomputed SNP gene weights to expedite the process. These SNP gene weights, as well as the SNP_gene_pair, are derived from the Genecode v46 GTF file, which is included in the quick_mode folder of the gsMap resource you downloaded. If you wish to use a custom GTF file, you can run the process step by step by specifying the --gtf_annotation_file parameter in the run_generate_ldscore command. For more detailed instructions, please refer to the documentation at: https://yanglab.westlake.edu.cn/gps_data/website_docs/html/step_by_step.html#generate-ldscore.

[Prakrithi-P]

Thank you! That makes sense. I missed the point that quick mode uses pre-defined weights. The step-by-step method worked!

Regards,
Prakrithi