luslab / arshamg-scrnaseq-wgan Goto Github PK

Hello,
I just read your manuscript Generative adversarial networks simulate gene expression and predict perturbations in single cells with great interest.
Congrats on the great work!
It seems that the code on this repo does not match the specifications of the paper though (classic GAN, rather than a Wasserstein GAN, different network architectures, optimizer...). I suspect this version of the code correspond to a former version of your paper.
Are you planning on updating it to match with the more recent manuscript?
Also, do you intend to provide the code to be able to reproduce the results? Notably, I'm interested about your algorithm to map the cells to the latent space, which, I suspect is critical to the results. If not, would you share some insights about the metric and the threshold you used to accept or reject the candidate simulated cells?

Hi Luscombe Group! Really neat work on this paper. I am just dipping my toes in the water with tensorflow and neural nets, but I'd like to try out your GAN implementation. Would you be willing to post your package and python versions, perhaps with a virtual environment and/or a requirements.txt file?

Thanks for sharing the code and data. I tried to install the git lfs and it said that "This repository is over its data quota. Account responsible for LFS bandwidth should purchase more data packs to restore access." If possible, would you mind provide an additional link for the "four_datasets_combined_lTPM_red_small_clean.csv" to run the code. Many thanks!