Pipeline Step 1 — clinically derisk human- and AI-generated molecular and genetic-variant predictions before any downstream design work.
What It Is
Genetic Variants is the leftmost stage of the BioinvestGPT pipeline — Variant Discovery, Step 1 on the Hub. It lets you clinically derisk human- and AI-generated molecular predictions across coding and regulatory genetic variants before you commit them to a hypothesis, a target program, or any later-stage work.
The catalog is backed by EVEE — Explaining Variant Effects with Embeddings — a named-biology variant-effect model. EVEE attaches interpretable disruption predictions to ClinVar variants, so instead of a single opaque "pathogenic" flag you see which biological functions a variant is predicted to disrupt and by how much.
The ClinVar variant identifier (e.g. chr17:81812181:G:A), with the short protein change (HGVS p.) underneath when available.
Gene
The gene symbol the variant falls in (e.g. TP53, BRCA1, GCGR).
Consequence
The molecular consequence (e.g. missense, nonsense, splice).
Significance
The ClinVar clinical significance label — Pathogenic, Likely Pathogenic, VUS, Conflicting, Likely Benign, or Benign.
Top disruption
EVEE's single highest-magnitude named-biology probe for that variant.
Score
The EVEE pathogenicity score (0–1), used to rank rows.
Where this fits: Variants are the raw molecular evidence at the front of the pipeline. Once selected, they flow into the Generator as voxels and feed a hypothesis, which you then carry forward to FIC Targets and the later BVCT mission stages. See Building & Confirming Hypotheses for what happens after the handoff.
You reach the catalog from the Hub. It is the first card in the pipeline timeline, to the left of FIC Targets.
1
Click the "Genetic Variants" card on the Hub
On the Hub, find the rose Variant Discovery card titled "Clinically Derisk Genetic Variants" (the leftmost / Step 1 card, marked with the DNA icon) and click Start.
The platform opens the Generator and immediately brings up the Clinically Derisk Genetic Variants catalog modal on top of the 3D knowledge graph. The modal header notes the source: 4.25M ClinVar variants with EVEE named-biology disruption predictions.
Tip: The card sets the Generator to open straight into the Variant Catalog, so you do not need to find it manually inside the Generator UI.
The catalog modal has three filters across the top. They re-query as you change them and reset to page 1.
Filter
How to use it
Gene symbol
Type a gene to narrow the list (e.g. TP53, BRCA1). Input is upper-cased automatically and debounced as you type.
ClinVar significance
Toggle the significance pills. Pathogenic and Likely Pathogenic are on by default; VUS, Conflicting, Likely Benign, and Benign are off.
Min pathogenicity
Drag the slider (0–1) to require a minimum EVEE pathogenicity score. The current threshold shows in the label.
Results are paginated 50 rows at a time. Use Prev / Next at the bottom to move through pages.
Indication-aware by design: the Hub catalog only returns variants that carry a clinical disease association, because a BVCT needs a clinical indication to be meaningful. Variants with no linked disease are filtered out of this view.
Tip: if the table shows "Search failed", the EVEE data service has not finished starting. Wait a moment and adjust a filter to re-run the query. Closing the modal (the X, or clicking the dark backdrop) returns you to the 3D graph; re-open it from the Hub card.
Select Variants
Selecting a variant is a two-step funnel: pick the variant in the catalog, then add the specific disruption you care about to your hypothesis.
2
View a variant's disruptions
On any row, click "View disruptions". The catalog modal closes, the matching gene voxel is selected in the 3D scene, and the gene's details panel opens scrolled to the Genetic Variants (EVEE) section, with your chosen variant pre-expanded.
That section lists the top ClinVar variants for the gene, ranked by EVEE pathogenicity. Each variant expands to its top named-biology disruption predictions, each shown with a category badge, a signed magnitude bar (red = increase, blue = decrease), and a numeric value.
Tip: If the gene is not currently loaded in the 3D scene, the platform logs that the gene is not in the current graph rather than selecting a voxel — load or search the gene in the Generator first.
3
Add a disruption to your hypothesis
Expand the variant, find the disruption that matters for your biology, and click "Add to hypothesis". There is no up/down/skip prompt — the disruption is added immediately as variant evidence.
This does three things on the gene voxel:
Routes the gene into the Variant group in the Generator's left sidebar (under the Hypothesis tab) — one row per added disruption.
If the variant maps to a ClinVar disease, adds the matching disease voxel(s) and triggers the radiating graph layout, exactly like the "Select Disease(s)" flow.
Appends one structured [EVEE] evidence annotation to the gene voxel (described below).
Tip: You can add several disruptions across several variants and genes. They accumulate — nothing is shipped until you confirm the hypothesis.
The [EVEE] evidence annotation
Each disruption you add becomes one annotation on the gene voxel, prefixed with [EVEE] so it is both human-readable and machine-parseable. It records the variant, gene, protein change, consequence, the named-biology probe, the disruption category, magnitude and value, the ClinVar significance, and the linked disease. A representative entry reads:
Field
Example
prefix
[EVEE]
mode
variant-evidence
variant / gene / protein
chr17:81812181:G:A · GCGR · p.…
probe / category
the named-biology probe name and its category badge
magnitude / value
signed disruption strength (e.g. value=+0.412)
clinvar / disease
the ClinVar significance and linked disease label
Variant evidence carries no regulation direction of its own (the mode is variant-evidence). If you want to set an up- or down-regulation direction on the gene, do it separately with the voxel's regulation control — that lives on the voxel, not inside the [EVEE] string.
Tip: each [EVEE] entry is its own removable row in the sidebar's Variant group. Removing the last [EVEE] entry on a voxel automatically clears its variant-evidence marker, so you can prune evidence without leaving stray flags behind.
From Variants to a Hypothesis
Added disruptions accumulate into the in-progress hypothesis you are assembling in the Generator. The variant step does not auto-ship anything and does not navigate away — you stay in the Generator and keep building.
4
Review the Variant group in the left sidebar
Open the Generator's left sidebar and switch to the Hypothesis tab. Your added disruptions appear in the Variant group, one row per [EVEE] annotation, each showing the gene, variant, and disruption it represents. Any disease voxels the variants mapped to are reflected in the graph's radiating layout.
Tip: The variant rows live alongside whatever else you have regulated. A single hypothesis can mix variant evidence with directly up- or down-regulated nodes and edges.
5
Confirm and hand off to the Validator
When the hypothesis is complete, click "Confirm All Regulations". This is the single action that packages the assembled hypothesis — including your variant evidence — and carries it forward into the Validator to design a BioinvestGPT Virtual Clinical Trial (BVCT).
From there the workflow joins the rest of the pipeline: the Validator stratifies patients, generates protocols, and predicts effect sizes leading to a GO / NO-GO decision.
Tip: Because the [EVEE] annotations are structured, the variant evidence travels with the hypothesis rather than being lost at the handoff.
Hypothesis (with variant evidence) handed to the Validator
Do I have to start from a variant?
No. Variants are the optional, earliest entry point. You can also begin a hypothesis from FIC Targets and never open the catalog. The variant card simply lets you ground a hypothesis in concrete, clinically-derisked molecular evidence first.
Can I add more than one disruption?
Yes. Add as many disruptions across as many variants and genes as your hypothesis needs — they accumulate until you confirm.
What if "Add to hypothesis" is disabled?
That action requires the Generator to be in interactive mode. If it is greyed out, the section shows a note explaining the Generator is not ready; reopen the Generator from the Hub card and try again.
Continue to Building & Confirming Hypotheses for the full hypothesis-construction flow, or to The 3D Knowledge Graph for navigating and selecting voxels. Prospective track record: data.bioinvestgpt.com.
BioinvestGPT BVCT Platform User Guide — Clinically Derisk Genetic Variants (EVEE). BVCT outputs are model-based decision-support analyses, not investment advice. Prospective track record: data.bioinvestgpt.com.
