Community Prediction Competition –

Myllia | Echoes of Silenced Genes: A Cell Challenge

Are you ready to participate?

Echoes of Silenced Genes: A Cell Challenge | Kaggle

Modelling cell behavior in response to stimuli and perturbations has been an increasingly popular topic. The so-called virtual cell models promise to navigate, or even replace lab experiments and thus significantly contribute to advances in medical research. However, we are yet to see such a capable model. In fact, we are yet to see a definition of what a good model is – the evaluation of prediction quality itself is not a simple task.

Single-cell RNA-seq has been used to catalog mRNA expression levels of almost all accessible cell types. Summarized in large atlases like CELLxGENE or scBaseCount, this type of data attracts data scientists eager to mine it for biological insights. While the bulk of these datasets contains only information on steady-state cells, the advent of techniques like CROP-Seq and Perturb-Seq has enabled the generation of very large datasets containing transcriptomes of cells after a CRISPR perturbation.

The increasing abundance of datasets naturally triggers ideas like: Having seen what the perturbation does in one cell type, can we predict what it will do in another? Your objective in this challenge is to predict how a human cancer cell line responds to perturbations induced by CRISPR interference (CRISPRi).

We organize this challenge to:

➜ See how capable current models of cell responses are
➜ Evaluate a new metric of prediction quality

📍 Join us via Echoes of Silenced Genes: A Cell Challenge | Kaggle
🗓️ Closing date of the competition: March 16, 2026