Add new method: LINGER

[emvg]

Integration of a New Method: LINGER

LIfelong Neural network for GEne Regulation : LINGER

Overview

We added :

• A rule to download prior knowledge used by LINGER (pre-trained networks).
• A definition file for Singularity (I did not use the dwn_image rule, but I can upload the ready .sif image to Zenodo if needed)
• The relevant files to support the parallelized version of LINGER (see Linger Parallelization Durenlab/LINGER#75)
• Original LINGER implementation with final cell population GRNs produced from the three LINGER outputs : TF binding potential, cis-regulatory network and trans-regulatory network (Details)

[PauBadiaM]

Hi @emvg,

Sorry for the slow reply, I was tied up with other projects. This is great work, thanks a lot for putting it together. LINGER was high on my list to include and I never had the bandwidth, so this is really appreciated.

The implementation itself looks solid. The config registration, the output naming, the linger_prior download, and the scripts are all in place. The main thing still missing is the generation of the per-method prior knowledge, specifically the TSS annotation.

Each method needs its own dbs/hg38/gen/tss/<method>.bed.gz, produced by a gen_tss_<method> rule. We build the method list for the TSS analysis automatically from the config, so as soon as LINGER is registered the genomic evaluation will look for dbs/hg38/gen/tss/linger.bed.gz and fail without it. You can see how the others do it in:

• workflow/rules/dbs/tss.smk (add a gen_tss_linger rule, plus the matching entry in the localrules: line)
• workflow/scripts/dbs/gen/tss/ (one script per method)

The bed should reflect the gene/TSS annotation LINGER actually uses, so the genomic comparison stays fair.
Once the TSS rule is in, the method should flow through the full benchmark. Happy to help if anything about that step is unclear.

And yes, with this added we would be glad to discuss including you as an author, since the paper is still in revision. Thanks again, this is a really nice contribution! 😄