search

andreyshabalin / MatrixEQTL

Matrix eQTL: Ultra fast eQTL analysis via large matrix operations
53 stars 16 forks source link

Help understanding option 'noFDRsaveMemory' in Matrix eQTL #14

Closed sharvari14 closed 3 years ago

sharvari14 commented 3 years ago

Dear Dr.Shabalin,

I am writing to seek your help in understanding the 'noFDRsaveMemory' option in Matrix eQTL function. Using this option as FALSE and TRUE, results in outputs with significantly different raw p-values. For example, setting this option as FALSE, returns p-values as low as e-300, while TRUE returns p-values ~e-15

I've read the paper and looked at the GitHub page, but couldn't get a better understanding of this behaviour. I would really appreciate it if you could shed some light on this, or point me to literature that can help to explain this behaviour.

Thank you

andreyshabalin commented 3 years ago

Dear @sharvari14,

The results should be the same regardless of the noFDRsaveMemory parameter, but with it set to TRUE the output is not sorted by p-value. So the top of the file would look different.

Do you actually observe different p-values for the same gene-SNP pair?

If so, can you provide a reproducible example I can run?

Thank you,

Andrey

gujjas commented 3 years ago

Hi Andrey,

Thank you so much for the reply. You are right. The p-values are the same for the same gene-SNP pair. Sorry about that.

I would like to confirm if it's expected to get very small p-values/FDR from this tool? SNP gene beta t-stat p-value FDR snp1 gene1 Inf 2.2250738585072e-308 1.08110553592833e-302 snp2 gene2 0.547772328108398 392.623084826711 2.2250738585072e-308 1.08110553592833e-302 snp2 gene2 0.547772328108398 392.623084826711 2.2250738585072e-308 1.08110553592833e-302 snp3 gene3 0.566147507210239 303.377306552224 2.2250738585072e-308 1.08110553592833e-302

Appreciate all the help. Thank you.