BioinvestGPT ApS · Research use only

Platform · Genetic variants · 8 Jul 2026

Clinically Derisk Every Genetic Variant Before Preclinical Spend.

BVCT sealed a FAILURE call on Verve’s VERVE-101 base editor (No. 1306) on 31 Oct 2023; Verve’s first human data followed on 12 Nov 2023, and the dashboard classifies that readout as True Negative (TN). The same engine called Intellia’s nex-z (No. 1116, True Negative (TN)) and Ionis’ olezarsen (No. 1406, True Positive (TP)).

BioinvestGPT ApS is not affiliated with, endorsed by or partnered with the other AI companies named on this page.

Coding or regulatory. Human-curated or AI-generated. Clinical impact vs SoC per variant, before a single dollar of wet-lab.

Every genetics-anchored programme — from gene therapy to base / prime editing, from antisense to siRNA, from biomarker stratification to in vivo CAR-T — starts with the same problem: a shortlist of candidate variants whose clinical consequence is unknown until the trial reads out. gnomAD tells you what is rare, and ClinVar how labs have classified it clinically. AlphaMissense and ESM-Variant tell you what is likely damaging. None of them tells you which variant, against which standard of care, in which indication, in which patient stratum, will deliver a registrational-endpoint hazard/odds/relative ratio below 0.6. That is the question that decides clinical capital.

BVCT now answers that question variant by variant. The Clinically Derisk Genetic Variants capability is live at bvct.bioinvestgpt.com. It clinically derisks human-curated and AI-generated molecular predictions across coding and regulatory genetic variants — with no patient data ingested, no ML training on past trial outcomes, and with a per-variant mechanistic causal rationale.

The atoms differ. The variant differs. The clinical outcome differs. BVCT closes the loop — one variant at a time, before the lab spends a dollar.

What the Variant-Derisking Loop Returns.

For every candidate variant — SNV, indel, CNV, splice-site, promoter, enhancer, UTR — BVCT returns four outputs at ~1 week per variant, < 0.1% the cost of a head-to-head Phase-III:

• Simulated Phase-III HR / OR / RR vs. Standard of Care at 0.1 precision, per indication and per patient stratum the sponsor specifies.
• GO / NO-GO call against the sponsor’s chosen clinical effect size threshold (e.g. HR < 0.6 vs SoC as the blockbuster threshold).
• Mechanistic causal rationale tracing every prediction to the root cause that drives it — the variant’s effect on expression, on the protein, on the pathway, or on disease biology.
• Patient-stratification hypotheses that maximise the head-to-head effect size before a trial protocol is locked, if applicable.

Inputs are structured and bounded: variant coordinates plus the sponsor’s specified indication and comparator standard of care.

Why This Matters for Decision Makers

Variant-anchored programmes combine biological elegance with high clinical risk in modern pharma R&D: genetic support makes a programme about 2.6 times more likely to succeed (Minikel et al., Nature, 2024), yet only about 10% of clinical programmes reach approval. In our estimate, a genetic-variant-based asset today commits nine figures of clinical capital against a candidate whose Phase-III readout depends on the specific functional consequence of a specific change at a specific locus — a question that wet-lab work cannot fully answer.

BVCT inserts a deterministic clinical-derisking layer between the variant nomination step and the IND-enabling commit. For a portfolio committee, that means: every genetics-anchored asset that crosses the GO line into clinical spend has carried a per-variant Phase-III hazard/odds/relative-ratio call, with a mechanistic rationale.

The only question that matters is which variant beats standard of care. BVCT answers it before the lab opens the freezer.

BVCT is Already Prospectively Validated.

The Variant-Derisking module runs on the same BVCT engine whose prospective record is now in public view: 705 validated ex-ante prospective clinical predictions — 95.6% accuracy, 89.7% PPV (GO calls), 99.1% NPV (NO-GO calls), F1 of 93.8% — against a 28.9% industry baseline (BIO / QLS / Informa 2011–2020, Phase II transition success rate). That record includes direct gene-editing and nucleic-acid-therapy predictions — see a prime example below.

Prime Example · True Negative · NO-GO Confirmed by the Readout · Verve Paused Enrolment in 2024

No. 1306 · Verve Therapeutics · VERVE-101 · In vivo Base Editing · Heterozygous Familial Hypercholesterolemia / ASCVD · Phase 1 · NCT05398029 · Heart-1

LNP-delivered adenine base editor (ABE mRNA + optimised gRNA) effecting a permanent A·T→G·C edit at the PCSK9 locus — designed to permanently silence hepatic PCSK9 expression and lower LDL-C in a single lifetime infusion for HeFH patients with established ASCVD. Sealed 31 Oct 2023: BVCT called FAILURE — commercially insufficient clinical benefit vs. SoC PCSK9 inhibitor evolocumab.

12 days later, on 12 Nov 2023, Verve announced the first-ever human proof-of-concept for in vivo base editing: VERVE-101 achieved blood PCSK9 protein reductions of up to 84%, LDL-C reductions of up to 55% and grade 3 TEAEs, all of which are inferior to the SoC evolocumab, which achieved near-complete PCSK9 suppression and LDL-C reductions of about 60% with mostly minor injection-site reactions (doi: 10.31083/j.rcm2505190). BVCT’s virtual cardiovascular patients had traced what the early pharmacodynamic data could not reveal: the permanent A·T→G·C edit at the PCSK9 locus would prove insufficient to clear the commercial bar against an established, well-tolerated PCSK9 inhibitor standard of care — the dashboard classifies that readout as True Negative (TN). Verve itself paused Heart-1 enrolment on 2 Apr 2024; that was not an FDA clinical hold: the FDA’s hold on the VERVE-101 IND ran from November 2022 to 23 Oct 2023, before the call was sealed.

Further Genetic Variant & Nucleic-Acid Programme EX-ANTE Calls

Derisk your next genetic variant. Before the lab opens the freezer.

Pick a candidate variant — SNV, indel, CNV, splice-site, or regulatory — and receive a simulated Phase-III clinical effect size vs. your specified standard of care within one week. No patient data required. No model training. A causal rationale you can defend.

Clinical Foresight Before Decisions. Earlier is Kinder for Patients.

Start at bvct.bioinvestgpt.com →