Documentation
Getting started, key concepts, and where the data comes from.
From sequence to sourced report in three steps
No installs, no pipelines to wire up. Load a target and start discovering.
Import your target & protein
Load a protein by UniProt ID, AlphaFold ID, or FASTA sequence and inspect it in 3D instantly.
Run the analyses
Docking, ADMET, PK/PD, response simulation and safety signals — run them side by side and rank candidates with calibrated confidence.
Generate an evidence report
Produce a restrained, research-use AI report with sources, assumptions and limitations, and share it with your team.
Key concepts
What is a program?
A program is a workspace for one discovery effort — its targets, candidates, analyses and reports live together and move through the 12-stage pipeline.
How does the job queue work?
Long analyses run as background jobs. A job moves through queued → running → succeeded, or ends in a labelled failure (dead-letter) after retries; the UI polls until it reaches a terminal state.
What does pLDDT mean?
pLDDT is a per-residue model-confidence score from 0–100 (blue = very high, red = very low). It reflects how sure the model is about local structure, not experimental accuracy.
Which sequence lengths can be folded?
The public ESMFold API handles sequences up to ~400 residues. For longer proteins, supply a UniProt accession to retrieve an AlphaFold model of any length.
Is my data used to train models?
No. Your sequences and results are used only to produce your analyses and are never fed into model training.
Data source references
The public databases ProteinInsight reads from. Each opens the source's official site.
Protein sequence and functional annotation.
Experimental 3D structures (X-ray, cryo-EM, NMR).
AlphaFold predicted structure database.
ESMFold single-sequence structure prediction.
Bioactivity and medicinal chemistry database.
FDA open data, including FAERS adverse events.
This documentation summarises current behaviour and may change as the platform evolves.