add Benchmarking a foundational cell model for post-perturbation RNAseq prediction

[szalata]

https://www.biorxiv.org/content/biorxiv/early/2024/10/01/2024.09.30.615843.full.pdf

[geroldcsendes]

Hi, thanks for your interest in our research!

In summary, we found:

• Simple baseline models can outperform scGPT on perturbational downstream tasks.
• Incorporating prior knowledge features, such as GO terms of perturbed proteins, can improve the performance of baseline models.
• The most widely used benchmarking datasets contain significant biases, making them suboptimal for evaluation.
• Models do not appear to benefit significantly from single-cell data, as those trained on bulk data can match or even outperform models trained on single-cell data.

Let me know if you need further information from us.

[Paulos2411]

Hi Gerold,

thanks for your summary. Could you perhaps summarize the information in a way as it has been done on the website for your specific case.

https://theislab.github.io/single-cell-transformer-papers/

I will then add them or feel free to create a PR.

Thanks :)

[geroldcsendes]

Hi, I made the PR here: #74

[Paulos2411]

Great, thank you.