Author a target-disease hypothesis in the Generator, then hand it off to the Validator for BVCT protocol generation.
Start From an Example or From Scratch
The Generator is where you author the biological hypothesis that a BVCT will validate: a disease, the targets you believe drive it, and the direction (up or down) in which each target is dysregulated. You can start two ways — load a curated bold hypothesis with one click, or build your own around a disease you choose.
2
Open the Load Example(s) popup
When you open the Generator, the Load Example(s) popup appears automatically with 13 curated bold hypotheses spanning seven therapeutic areas (Oncology, Metabolic, Neurology, Immunology, Cardiovascular, Rare disease, Aging / Fibrosis). You have two paths forward:
Load a bold example (recommended) — Pick a card and the disease, regulated targets, and rationale annotation all load together. Cards tagged Enhancement-ready also seed the drug-design brief (seed drug, modality, target protein, developability goals) so the downstream ENHANCEMENT BVCT workflow is pre-populated.
Start from scratch — Close the popup (X or Escape), then click Select Disease(s) in the top toolbar and type your disease name (e.g., "non-small cell lung cancer"). The graph rebuilds around that disease with no pre-regulated nodes — the TARGETS tab starts empty so you can author your own hypothesis node-by-node.
Tip: Picking a new example while one is already loaded clears the previous example's regulated nodes and annotations first — no manual cleanup needed.
3
Narrow the list with therapeutic-area filters
Use the therapeutic-area filter chips at the top of the popup to narrow the 13 examples to a single area, or click Randomize in the header to pick one uniformly at random within the current filter. This is the fastest way to see a complete, regulated hypothesis end-to-end before authoring your own.
4
Confirm the example loaded onto the graph
Once an example loads, the 3D knowledge graph centers on the example's disease and its regulated targets appear with green (up) and red (down) pills already applied. The left sidebar TARGETS tab lists every regulated node and edge. From here you can confirm as-is, or continue annotating to extend the hypothesis (see Annotate targets with regulation).
Tip: If you started from scratch instead, this view shows your chosen disease at the center with its neighborhood radiating outward and an empty TARGETS tab — ready for you to add the first regulated node.
Knowledge graph: The Generator is powered by PrimeKG — roughly 130,000 biomedical entities and 8 million curated relationships. The radiating layout shows your disease at the center with associated genes, proteins, drugs, and pathways in concentric layers; each layer is one hop in the graph.
Load a Blockbuster Drug (BIC Drugs)
Authoring a target hypothesis is the FIC (First-In-Class) path. If your goal is instead to optimize an existing drug — a Best-In-Class (BIC) move — the Hub offers a separate entry point.
The Hub's BIC Drugs card opens a dedicated Load Blockbuster(s) modal — separate from the FIC Targets Load Example(s) flow described above. The modal lists curated blockbuster drugs across modalities (semaglutide, pembrolizumab, trastuzumab deruxtecan, sotorasib, and more) with revenue, indication, and modality metadata.
Pick a blockbuster, click Load, and the Validator opens with the drug pre-filled as your seed: SMILES (or sequence / HELM), modality, target protein, and current PK profile are all populated. From there you set differential design objectives and modality-specific levers in the ENHANCEMENT workflow.
Tip: Loading a blockbuster is the fast path to "give me a me-better candidate to a known drug" — typically the first move when scouting BIC opportunities for lifecycle extension or competitive entry.
Annotate Targets With Regulation
The core of a hypothesis is encoding which targets you believe are dysregulated, and in which direction. You do this by right-clicking nodes on the 3D graph and assigning green or red pills.
5
Right-click a target to up- or down-regulate it
Right-click a target gene and select "Effective Up-Regulate" (green pill) or "Effective Down-Regulate" (red pill). This encodes your biological hypothesis:
Green pill = You believe this target is activated / upregulated in the disease
Red pill = You believe this target is suppressed / downregulated
These annotations travel with the hypothesis into the BVCT protocol design.
If you loaded an example, the target is already regulated for you — this step is for users who started from Select Disease(s), or who want to add more regulated nodes on top of an example.
Pill
Meaning
When to use
Green / Up
Target is activated or over-expressed in the disease
Hypotheses about driver oncogenes, gain-of-function signaling, over-active pathways
Red / Down
Target is suppressed or under-expressed in the disease
Hypotheses about lost tumor suppressors, depleted protective factors, silenced pathways
Tip: You can mix green and red regulations across many nodes in a single hypothesis — the BVCT design accounts for the full set of regulated targets and edges together, not one at a time.
Review and Confirm Your Hypothesis
Before handing off to the Validator, review the complete hypothesis in the left sidebar and confirm it.
6
Review the hypothesis summary in the left sidebar
The left sidebar shows all your annotated nodes and edges. Review the hypothesis summary: disease, regulated nodes (with green / red badges), and edge count. When everything looks right, click "Confirm All Hypotheses".
Tip: If you loaded an example, the confirm step is one click — the example already contains a complete regulated hypothesis.
7
Confirm to package the hypothesis
Confirming packages your hypothesis — disease, regulated targets, and supporting literature — into a structured handoff. This is the moment the Generator stops being an exploration tool and produces a validated artifact ready for BVCT protocol generation.
Before you confirm: Make sure every node you intended to regulate actually shows a green or red pill in the sidebar. Nodes you only clicked or hovered are not part of the hypothesis — only explicitly up- or down-regulated nodes carry over.
Hand Off to the Validator
Confirming a hypothesis is also the bridge between modules. The platform carries every field over automatically — you do not re-enter the disease or targets.
8
Navigate to the Validator
After confirming, the platform packages your hypothesis (disease, regulated targets, supporting literature) and navigates you to the Validator for BVCT protocol generation. All data carries over automatically — the Validator opens with your disease and therapy already pre-filled, ready for you to set the BVCT Type and Purpose and begin patient stratification.
Once in the Validator you continue the workflow described in the BVCT chapters: configure the trial, generate patient strata, review eligibility criteria and protocol cards, and approve the protocols you want to transmit. You can open the Validator directly at any time at bvct.bioinvestgpt.com/#/validator/.
Round-trip: The handoff is two-way. From a BVCT result in the Validator you can use Visualize Rationale to jump back into the Generator with the trial drug, indication, and predicted effect size pre-populated — closing the loop between hypothesis and validated outcome.
BVCT outputs are model-based decision-support analyses, not investment advice. The prospective track record is published at data.bioinvestgpt.com.
Capturing the 3D Graph for Feedback
If something looks wrong on the 3D knowledge graph — a misclassified node, an edge that should not be there, a layout that obscures your hypothesis — you can capture the graph region and send it as feedback in two clicks.
Click the floating Feedback button at the bottom-right of any screen. The capture tool opens an overlay where you can draw on the snapshot using one of four marker colors (rose, amber, emerald, slate). Use Undo to remove the last stroke or Clear to start over. Strokes are composited into the PNG before sending — your annotated capture lands in the support team's inbox along with a description.
The capture works on the Three.js WebGL canvas (the 3D knowledge graph) as well as on regular DOM elements (sidebars, modals, results panels). No screen-recording permission is required.
9
Author your own hypothesis on a blank disease graph
When you start from scratch via Select Disease(s), the TARGETS tab begins empty and the graph radiates from your chosen disease with no pre-regulated nodes. This is the cleanest canvas for capturing feedback while you build: annotate a node, see how it lands in the sidebar, and if the graph misrepresents the biology, capture and annotate it with the Feedback tool right away.
Tip: Feedback captures are most useful when they include a short description of what you expected versus what the graph showed — the marker strokes point to where, the description explains why.
BioinvestGPT BVCT Platform User Guide — Building & Confirming Hypotheses. BVCT outputs are model-based decision-support analyses, not investment advice. Prospective track record: data.bioinvestgpt.com.
