Which model to use for genomic variants analysis?

[nc1m]

Dear authors,

first thank you for your work!
I'm trying to recreate the results from section "3.5 Identifying functional genetic variants with DNABERT".
Which model should be used to compute the predictions?
Did you fine-tune a model on functional variants or used a pre-trained (DNABERT6) or other fine-tuned model (e.g. DNABERT-TF, DNABERT-Prom-300 or DNABERT-Prom-scan)?

Thank you for your time.

[vitoriastavis]

Hello, did you figure it out? I'm working on a project about variants.

Also, what did you use for labels in the input for prediction? In DNABERT/examples/ft/6/ the files have 1 and 0, but I'm not sure about the meaning. Maybe it's 1 for a condition and 0 for control, or 1 for SNP and 0 for none?
And the file DNABERT/SNP/examples/dev.tsv has different labels, such as 23, 104, 125. These values look like IDs, not labels. Could anyone confirm?

Thanks in advance!