Computational Antifibrotic Research · in formation
Advanced decompensated cirrhosis with refractory ascites. Post-TIPS. Child-Pugh B/C. This is the population routinely excluded from modern fibrosis trials — and the population where even partial regression would be the single most meaningful clinical win in hepatology. Helix Fibrex Institute is being built to do the pre-experimental work that makes a credible, partner-validated research program for this population possible: computational hypothesis generation, structural grounding, and the long, cheap work that has to happen before anything becomes wet.
01 The focus
Can a partial reduction in hepatic fibrosis — not full reversal — meaningfully lower portal venous pressure enough to slow ascites reaccumulation in patients who already have TIPS in place?
Partial regression correlates with measurable portal pressure reduction: HVPG drops of 2–5 mmHg have been documented after hepatitis C cure in advanced cirrhosis cohorts. Full architectural normalization is rare; the pressure effect is not. The open question is whether interventions like TIMP-1 silencing, HDAC-mediated HSC quiescence reversion, FGF21 agonism (efruxifermin-class), or structured combinations thereof could achieve a comparable effect in established decompensated cirrhosis — and what the plausible dose-response profile looks like in a population whose hemodynamic and hepatic reserve constraints diverge sharply from the F2–F3 cohort that today's trials are built around. That is a question that starts in silico, not in the clinic.
Patient reality
Refractory ascites develops in 5–10% of cirrhotic patients once ascites appears, and carries one-year survival around 25% at historical baseline — a prognosis comparable to advanced-stage solid-organ cancers. No approved therapy addresses ongoing fibrosis progression specifically in the post-TIPS setting.
Mechanism
The MMP/TIMP imbalance that locks advanced fibrosis in place is kinetic, not thermodynamic. Cross-linked collagen matrix is disruptable given the right sequence of TIMP-1 inhibition, MMP reactivation, and hepatic stellate-cell deactivation.
Where computation fits
A credible hypothesis package for this population — ranked candidate compounds, structural rationale, a proposed path to organoid validation, and considerations for eventual trial design — is exactly the kind of pre-wet-lab output an AI-accelerated research pipeline can now produce at a fraction of historical cost. Producing that package is the institute's reason to exist; validation and trial stages live with partners.
02 The approach
The most expensive part of modern drug discovery is not hypothesis generation — it is the hypothesis-to-validation feedback loop. Helix Fibrex is being built around the pre-experimental portion of that loop: adversarial multi-agent reasoning over the literature (our Idea Folding methodology), grounded in PubMed-anchored retrieval, structural biology, and cheminformatics. The eventual output is hypothesis dossiers that a clinical or wet-lab partner could take into an organoid screen or, further out, into a trial proposal. Today the work is earlier than that: the institute is building the computational engine that would make such dossiers credible. It does not operate a lab, treat patients, or run trials — and is not positioned to.
Generators surface candidates from the full corpus. Falsifiers attack them with the explicit job of finding fatal flaws. Arbiters score survivors on mechanistic plausibility, novelty, and experimental accessibility. A persistent failure log ensures dead branches stay dead. This is the engine the institute is building first — the computational layer where hypothesis quality is won or lost before any other cost is incurred.
AlphaFold 3 structural models of TIMP-1 / MMP binding interfaces. Virtual docking against the FDA-approved compound library. Cross-referencing existing safety data to prefilter for translational feasibility in hepatic-reserve-limited populations. Still entirely computational; still entirely pre-wet-lab.
The near-term target: hypothesis dossiers packaged specifically for hepatic organoid validation, handed off to a wet-lab partner equipped to run fibrotic microHO assays on human tissue — without animal models or patient enrollment. Reaching a credible, partner-ready hand-off is a milestone the institute is funding-dependent to reach; it is not current operational capability.
Further out: contribution to PCORI-style patient-centered trial frameworks — proposals drafted with, not for, the refractory-ascites population. Trial design and execution are downstream partnerships the institute would contribute to rather than run, and are well beyond current scope. Everything past step 02 depends on the institute reaching funded operational maturity first.
03 Why now
2025
Akero's efruxifermin showed biopsy-confirmed reversal of cirrhosis in compensated MASH patients at 96 weeks — 39% of the 50 mg arm vs 15% on placebo (SYMMETRY Phase 2b; NEJM 2025, presented EASL and AASLD 2025). The field shifted from "cirrhosis is irreversible" to "partial regression is achievable in compensated patients." The decompensated cohort — Child-Pugh B/C, post-TIPS, refractory ascites — was excluded by design. That's the population this institute exists for.
2025
Google's AI co-scientist proposed epigenomic-modifier candidates that a Stanford group then tested against TGFβ-stressed human hepatic organoids; two drug classes — HDAC inhibitors (including the FDA-approved oncology agent vorinostat) and BRD4 inhibitors — showed measurable antifibrotic activity (Guan et al., Advanced Science 2025). Separately, a Stanford "Virtual Biotech" multi-agent framework produced a full target-assessment report for $46 in API costs (bioRxiv, 2026). The computational phase of this work stopped being capital-intensive.
2024–2026
Multi-lineage hepatic organoids with quantifiable fibrotic response (microHOs — JCI Insight 2024) are now a standardized capability at a growing number of groups. Drug screening on human tissue without animal models or patient enrollment has moved from frontier research to operational reality — for any computational group with the right wet-lab partner. The bridge from silico to human tissue did not exist two years ago.
04 Engage
For
Hepatologists and liver-group PIs working on decompensated cirrhosis, portal hypertension, or post-TIPS trajectories — we would welcome a conversation. Co-authorship, co-investigator positioning on grant architecture, and early access to the hypothesis dossiers are on offer.
For
AlphaFold applications, molecular docking, multi-agent orchestration, single-cell liver datasets. If you're building in any of those areas and fibrosis is within scope for you, there's room here.
For
Fiscal sponsorship in formation. The grant pathways we're orienting toward — AASLD Foundation Pilot Research Award, NIH R21, PCORI-aligned mechanisms — typically require a faculty co-investigator, and those partnerships are part of the institute's near-term architecture. Early donor engagement is welcome: the decompensated population does not have a lobby, and this is the moment when the smallest capital does the most work.
For
Lived experience is a structural asset in this kind of work, not a footnote. PCORI-weighted grants reward it. So does good research design.
Direct line
contact@fibrexinstitute.comBrief us in three paragraphs: who you are, what you're working on, how it overlaps. Every substantive message gets a response.
