Quick Start¶
This page covers the one-time setup — a short primer on the model, credentials, and install. For the full hands-on flow (data prep → DEGs → target validation → causal mechanism), follow the tutorial.
How TwinCell works¶
Most AI models in biology attempt the forward problem: given a perturbation, predict what happens to the cell. TwinCell solves the inverse problem — given an observed state change between two conditions, it reasons about the upstream protein regulators driving it. At its core the model performs target identification: scoring how causally each protein in a curated multi-omics interactome influences a transcriptomic state change.
You bring a target and one or two cell states, and TwinCell exposes that capability as two workflows:
| Use case | Input | Output | Status |
|---|---|---|---|
| Target validation | A target + one or two cell states (with DEGs) | Target efficacy — % of DEGs the target causally covers, plus the supporting causal paths / mechanism | Available |
| Simulation | A target + a control state | The predicted DEGs from perturbing the target | Coming soon |
Setup¶
Get your credentials¶
Sign in to the TwinCell console to create an account and generate an API key (keys start with dl_).
Open the TwinCell console to get your API key
Then set it in your environment (notebooks fall back to a getpass prompt, so your key is never echoed):
Install¶
Python 3.12+ is required.
Next steps¶
You're set up. The target-validation tutorial walks through the full flow end to end — load data, build pseudo-bulk profiles, compute the disease signature with PyDESeq2, then validate a target and inspect its causal mechanism.
- Run the tutorial: target validation (Deucravacitinib / psoriasis)
- Prepare your own data: TwinCell data requirements
- Browse the methods and parameters: API Reference
- Read the full methodology: TwinCell preprint