Don't Predict. Compute. · Dallas, TX

Map the disease.
Validate the drug.

Computational Disease Analysis reports for foundations, biotech, and pharma R&D.

PHYSIM maps disease failure architecture down to the cell, identifies structurally critical targets, and validates candidate engagement through a 7-gate deterministic gauntlet. No AI guessing. No confidence scores. Same molecule, same answer, every time.

6,675 diseases mapped. 97.8% concordance with clinical reality. Standard screening stops at pharmacokinetics — that's Gate 1. We go to Gate 7.

$2.0B
median R&D cost
12+ yrs
to approval
90%
failure rate

What if you could validate candidates, map disease engagement, and repurpose existing drugs — computationally — in weeks?

Sun, Zheng & Bhattacharya, JAMA Health Forum (2024) · BIO/QLS Clinical Development Success Rates (2021)
PHYSIM - Design. Simulate. Validate, Before the Lab.
PHYSIM Chamber — Live Platform
PHYSIM · The Chamber Live Platform
Not an LLM
I don't read papers about drugs. I compute the physics of survival.
Not Probabilistic
No confidence intervals. No training data. No probability. Physics.
Deterministic Computation
Same input, same result. Run it 10,000 times. Zero variance.
6,675
Diseases Mapped
97.8%
Clinical Concordance
254K
Patient Entities
7
Validation Gates

The difference between guessing and measuring

Every AI drug discovery platform on the market is probabilistic — it learns from past data, then predicts what might happen next. PHYSIM doesn't predict. It computes. That's not a marketing distinction. It's a physics distinction.

Probabilistic AI

Learns from the past.
Guesses the future.

1
Trained on historical outcomes
If a drug failed in 2019, the model learns "drugs like this fail." But it doesn't know why.
2
Outputs a confidence score
"78% likely to be orally bioavailable." What does the other 22% look like? Nobody knows.
3
Different answer each time
Run the same molecule twice, get two different scores. Retrain the model, get a third. Which one is real?
4
Blind to novel mechanisms
If it's never seen a mechanism before, it can't predict it. The training data is the ceiling.
The consequence
90% of drugs fail in clinical trials. The AI said they would work. It was guessing.
"AI has not demonstrably improved the ~90% clinical failure rate." — DeepCeutix, 2025
"We've just seen failure after failure" in AI drug discovery. — CEO, Deep Genomics ($250M), STAT News
Insilico's first end-to-end AI drug fell short on efficacy. Recursion's first AI-discovered drug showed no reportable efficacy. — STAT News, 2025
"We believe we can predict human responses from low-dimensional data with very little genetic diversity." — Nature
Deterministic Computation

Measures the physics.
Computes the answer.

1
Computes from molecular structure
The engine reads the molecule and simulates what happens when it enters the body. No history needed.
2
Outputs a measurement
"This molecule is destroyed by gastric acid at pH 1.2 because its structure breaks at the ester bond." Specific. Auditable.
3
Same answer every time
Run it on Monday or Friday. Run it on a laptop or a server. Same molecule, same computation, same result. Sixth decimal place.
4
Works on anything — including never-before-seen molecules
Physics doesn't need training data. If the molecule has a structure, the engine can compute what happens to it.
The result
97.8% concordance with clinical reality across 14,606 compounds — and that's from Gate 1 alone. The engine validates through 7 gates, down to the depth of a cell.
Drug development failure vs deterministic computation
Why does this matter?

Because a $2.6 billion drug failure starts with a confidence score someone believed.

The average cost of bringing a drug to market is $2.6 billion. Most of that cost is failure — compounds that AI platforms scored as "high confidence" that collapsed in human trials. The AI wasn't lying. It was doing what probability does: making its best guess based on pattern matching.

PHYSIM doesn't score confidence. It computes physics. When the engine says a molecule is destroyed by gastric acid, it means the structural computation showed bond-level failure at pH 1.2. That's not a guess you can disagree with. That's a measurement you can verify.

90%
Phase II failure rate
Nine out of ten drugs fail clinical trials — after years and millions invested.
97.8%
PHYSIM concordance
Matches clinical reality. Not a prediction. A measurement.
0
Variance between runs
Same molecule, same answer. Every time. On any machine. Forever.

When physics agrees with clinical reality

Every claim below was computed — not learned from training data, not inferred from patterns, not predicted from historical outcomes. The engine takes molecular structure as input and computes physical survival. These are the results.

The Suspension Bridge Principle

You don't predict that removing a load-bearing cable will collapse a bridge. You compute the tension forces, and the math proves the geometry cannot hold.

Standard computational biology uses pharmacokinetic (PK) screening — 15 independent pass/fail checkpoints. PHYSIM starts there, exceeds it, and then goes further: it introduces an entirely physically simulated human body down to the tissue level using proprietary techniques.

The molecule doesn't just pass a checklist — it survives the body.

Clinical Reality Gate 1 Gate 7 Bond Failure PK Tissue Cell Chronic DILI Disease
97.8%
Clinical Concordance
219,090 physical evaluations
Physics agreed with clinical reality across 14,606 compounds × 15 biological stages. Each cable is a validation gate. If one snaps — the molecule fails. No guessing.
❤️‍🔥

Cardiac Ion Channel Physics

The engine computes hERG channel pore geometry against molecular LogP and basic nitrogen structure. If the physical interaction produces channel blockade — the molecule is destroyed at the Cardiomyocyte Sarcolemma.

CONCORDANCE RESULT
657 compounds correctly killed by hERG physics — matching known cardiotoxicity profiles of withdrawn drugs like Terfenadine.
🧱

P-glycoprotein Efflux Physics

P-gp efflux pumps are the body's bouncers. The engine computes molecular substrate recognition — lipophilicity, aromatic ring count, and hydrogen bond donors — to determine if the molecule gets ejected before it can act.

CONCORDANCE RESULT
766 compounds correctly identified as efflux substrates — the #1 kill gate across the entire 14,606-compound gauntlet.
☠️

Withdrawn Drug Detection

609 compounds with known withdrawal histories were processed. Using only molecular structure — with no access to clinical outcome data — the engine independently computed the same failure mechanisms that led to their removal from market.

CONCORDANCE RESULT
100 of 609 withdrawn compounds independently killed by the physics. The remaining 509 passed — because some withdrawn drugs were pulled for non-pharmacological reasons (market, manufacturing), not because the molecule was physically unviable.
Concordance by Drug Classification

Every compound class in the global pharmacological registry — approved, experimental, investigational, withdrawn, and illicit — was computed through the full 15-stage gauntlet.

97.8%
APPROVED
2,237 drugs
98.2%
EXPERIMENTAL
7,919 compounds
97.1%
INVESTIGATIONAL
3,562 compounds
98.1%
WITHDRAWN
609 compounds
99.8%
ILLICIT
129 compounds

Per-stage concordance — the rate at which each individual biological computation agrees with observed outcomes. 14,606 compounds × 15 stages = 219,090 evaluations.

Beyond Drug Screening

Give us the disease. We compute the targets.

The engine doesn't just screen drugs — it identifies which molecular targets are structurally critical to a disease's survival. Remove them, and the disease architecture falls apart. This isn't pattern matching. It's the same principle that tells you which cable will bring down a bridge.

100%
Target Validation Rate
Every computationally identified critical gene, when removed, collapses the disease architecture. Verified across all tested diseases.
6,675
Diseases Mapped
Cancer, autoimmune, neurodegenerative, metabolic, cardiovascular, genetic, psychiatric, infectious, and rare diseases — all structurally analyzed.
0
Learned Weights
No training data. No model drift. No retraining. Same disease, same targets, same answer — every time. Deterministic by design.

The engine maps both halves: which targets hold the disease together, and which compounds can engage them. Drug screening and target discovery — from the same computation.

The 7-Gate Gauntlet

Every drug must survive
all seven gates.

Every other platform on the market stops at Gate 1 — standard ADMET pharmacokinetic screening. That's it. Gates 2 through 7 don't exist anywhere else. PHYSIM tests whether a drug survives the body, reaches the right tissue, penetrates the cell, remains safe across dosage and duration, avoids liver destruction, structurally engages the disease, and matches the patient's stage. Seven gates. Zero shortcuts.

1
Does the drug survive the body? (industry standard — everyone stops here)
2-3
Does it reach the tissue and penetrate the cell?
4-5
Is it safe across dosage and duration? Does it damage the liver?
6-7
Does it engage the disease — and can this patient respond?

The entire industry stops at Gate 1. PHYSIM is the only engine that goes from pharmacokinetic survival all the way to patient-specific, stage-dependent treatment matching. Gates 2 through 7 are proprietary. No one else has them.

PHYSIM 7-Gate Gauntlet — Drug Validation Funnel
McCauley Convergence Cascade — 7-Stage Universal Disease Staging
Universal Disease Staging

Every disease follows
the same structural collapse.

We tested 9 completely different disease cascades — cancer, autoimmune, neurodegenerative, genetic. Every single one shows the same mathematical pattern: structural connectivity decreases monotonically as the disease progresses. Stage by stage. The network falls apart the same way every time.

9 / 9
Cascades confirmed
6,675
Diseases mapped

This means you can know where a patient is in their disease progression — mathematically. And you can compute which drugs work at which stages.

Today, every disease has its own staging system — none of them talk to each other. This is one mathematical function that stages every disease on the same scale. Objective. Universal. Deterministic.

Where disease architecture becomes drug targets.

My proprietary platform translates raw biological failure into actionable, targetable molecular blueprints. The engine maps the full structural progression of a disease — from first cellular compromise through irreversible collapse — and identifies exactly where a drug can intervene.

PHYSIM — The Chamber: 15-Stage Biological Gauntlet Simulation Environment
PHYSIM · The Chamber
15-Stage Biological Gauntlet · Real-Time ADMET Validation
Live Platform Screenshot

Watch a compound fight for survival.

The interactive ADMET gauntlet below demonstrates how a molecule is tested against each biological checkpoint in real time.

One deterministic engine. Four commercial outcomes.

Disease architecture, critical target computation, candidate validation, and pre-clinical reporting — computed as one structural problem.

🧬

Disease Architecture Mapping

Give us a disease. We return its structural blueprint — every pathway, every convergence point, every stage of collapse from first cellular compromise through irreversible progression. We don't read the literature. We compute the architecture. 6,675 diseases mapped and counting.

🎯

Target Acquisition & Drug Repurposing

The engine identifies which molecular structures can engage a disease — including FDA-approved drugs that may already work but have never been tested against the target. New candidates. Existing drugs. All computationally validated against the disease's actual failure architecture.

⚗️

7-Gate Candidate Validation

Every candidate survives a 7-gate gauntlet — from pharmacokinetic survival through tissue penetration, organelle-depth interaction, chronic dosage safety, liver toxicity, disease engagement, and patient-stage matching. The molecule doesn't pass a checklist. It survives a computational body.

🛡️

ADMET Is Gate 1. We Go to Gate 7.

Traditional tools stop at pharmacokinetics — absorption, distribution, metabolism, elimination, and toxicity. That is the first gate inside a much larger computation. PHYSIM continues through tissue penetration, organelle-depth interaction, chronic exposure, hepatotoxicity analysis, disease engagement, and patient-stage matching. ADMET is included. It is not the category.

📋

Pre-Clinical Grade Report

You receive a complete, auditable report documenting every gate: pharmacokinetic profile, safety panel (hERG, BBB, CYP450), chronic toxicity timeline across dosage and duration, DILI proximity scoring, disease engagement mapping, and patient stage classification. Ready for your wet lab, your investors, or your regulatory filing.

Computed against clinical reality — and verified

These are not projections. Every metric was earned through 219,090 individual physical evaluations — benchmarked against known clinical outcomes across the full pharmacological registry. No model. No training data. Pure structural physics.

🔬
97.8%
Clinical Concordance

219,090 individual physical evaluations against 14,606 compounds. Not predictions — deterministic computations from molecular structure. Each of 15 pharmacokinetic stages was independently scored against clinical outcome data.

🛡️
97.8%
Elevated ADMET Concordance

The Synthetic Human Chamber — a proprietary body simulation engine — correctly computes the physical survival of 97.8% of FDA-approved compounds through 15 physiological kill gates. The 2.2% gap is the frontier of approximation, not prediction error.

🧫
14,606
Compounds Computed

Every compound with a valid molecular structure in the global pharmacological registry — approved, experimental, investigational, and withdrawn — was processed through the full engine at 215 compounds per second. No model. No training data. Pure structural physics.

ADMET is a checklist. This engine simulates a body.

Standard computational screening treats each biological checkpoint as an independent pass/fail gate — a checklist. The Synthetic Human Chamber simulates continuous body traversal: a molecule enters the GI tract, survives absorption, distributes through real tissue geometry, and either reaches the target or gets killed. Compounds that traditional screening incorrectly rejects are recovered. The ones that are genuinely dangerous are caught.

97.85%
Per-Stage Concordance
Across 219,090 individual physical evaluations. Each of 15 biological stages independently verified against clinical outcomes.
2,922
Compounds Correctly Killed
Drugs the physics determined could not physically survive the human body. Each kill traced to a specific biological stage with a provable mechanism.
766
P-gp Efflux Kills
The single deadliest biological gate. The body's own efflux pumps ejecting compounds before they can act — computed from molecular structure alone.
Computed Against Withdrawn Compound Profiles

The system was tested against 609 compounds with known withdrawal histories. Using only molecular structure — with no access to clinical outcome data — the engine independently computed the same failure mechanisms that led to their withdrawal.

Metabolic & Cardiac Destruction
Primary Kill Mechanism — 47% of Kills

hERG cardiac channel blockade, metabolic instability, and short half-life — compounds the body destroys or that stop the heart before reaching the target.

Absorption Barriers
Secondary Kill Mechanism — 39% of Kills

P-glycoprotein efflux (766 kills), gastric acid degradation (278 kills), and aqueous insolubility (101 kills) — compounds the body actively rejects or cannot dissolve.

Hepatotoxicity Screening
Tertiary Kill Mechanism — 14% of Kills

405 compounds flagged for liver toxicity risk — including structural alerts, lipophilicity-driven hepatic extraction, and reactive metabolite formation. Standard screening misses 85% of these.

Validated Across All Drug Classes
14,606
Compounds Tested
215/sec
Processing Speed
610
Withdrawn Drugs Tested
100
Withdrawn Drugs Killed

Not a faster version of the old workflow. A different computational category.

Traditional pharma fragments the problem. Standard AI accelerates fragments. PHYSIM computes the full structural relationship between disease architecture and candidate survival.

Discovery Protocol
Traditional Pharma
Standard AI Biotech
The PHYSIM Engine
Disease Context & Mapping
Isolated target testing based on established literature.
Single-pathway analysis using off-the-shelf ML models.
McCauley Convergence Cascade. Cross-references 6,675 diseases to find upstream control points.
Drug Evaluation Depth
Discovered in late-stage animal/wet lab testing.
Basic tox-screening and structural prediction. ADMET-level.
7-gate deterministic gauntlet. ADMET at Gate 1, then tissue, organelle, chronic, hepatotoxicity, disease engagement, and patient-stage fit.
Time to Candidate
4 to 6 Years.
6 to 12 Months.
Weeks. Delivered ready for the wet lab.
Cost Profile (Pre-Clinical)
$50M - $100M+
$5M - $15M
Competitively priced engagements. Predictable, project-scoped.
Methodology
Wet lab screening. Statistical survival analysis.
ML predictions from training data. Probabilistic confidence scores.
Deterministic physics computation. No training data. No model. No probability.
Engagement Model
Internal teams + outsourced CROs. Multi-year contracts.
SaaS platform license. You run the models yourself.
Full-service delivery. I run the engine — you receive validated candidates and reports.

We replace years of screening with weeks of computation

Traditional drug discovery costs $50–100M and takes 10–15 years before reaching a single patient. PHYSIM compresses the entire pre-clinical pipeline into a deterministic computational engine — delivering wet-lab-ready candidates in weeks.

10–15 yrs
Traditional Timeline
→ Weeks with PHYSIM
$50–100M
Traditional Sunk Cost
→ Fraction with PHYSIM
90%
Phase II Failure Rate
→ De-risked computationally
⚠️
Why AI Drug Discovery Keeps Failing

Most AI startups generate molecular candidates using statistical probability (LLMs, GANs). The human body doesn't obey statistics — it obeys physics. PHYSIM is deterministic: it maps the physical survival thresholds a compound must pass before it can exist inside a living human. Nothing is generated until the structural physics says it can survive. This is why our engine flagged the lethal drug Cisapride before the FDA withdrew it.

Traditional step
Proprietary IP
Partner stage
TRADITIONAL PIPELINE

Disease selection

6–12 months

Target identification

2–4 years • $5–10M

High-throughput screening

1–2 years • $2–5M

Hit-to-lead optimisation

1–3 years • $5–15M

ADMET / pharmacokinetics

1–2 years • $3–8M

90% failure rate at Phase II

Pre-clinical animal testing

2–3 years • $10–30M

IND filling & regulatory

6–12 months • $1–3M

Phase I clinical trial

1–2 years • $10–20M

TRADITIONAL TOTAL

10–15 years • $50–100M

PHYSIM ENGINE

Genomic architecture engine

Disease signal detection

Proprietary

Disease convergence mapping

Cascade & pathway resolution

Proprietary

Target acquisition

6,675 diseases • 7-gate engine

Proprietary

Molecular design engine

PM-class candidate generation

Proprietary

PHYSIM synthetic human

14,606 compounds · 97.8% concordance

Proprietary

7-Gate gauntlet

PK · Tissue · Organelle · Chronic · DILI · Disease · Stage

Proprietary

PHYSIM TOTAL

Weeks • fraction of the cost

OUTPUT

Candidate output

Validated • ready for wet lab

Proprietary

Wet lab validation

Partner opportunity

WHAT THIS REPLACES

  • Years of screening
  • Millions in sunk cost
  • 90% Phase II failure rate
  • Animal testing overhead
  • IND filing uncertainty

WHAT YOU RECEIVE

  • Validated candidates
  • Full 7-gate biological report
  • Disease engagement mapping
  • 97.8% ADMET concordance benchmark
  • Ready for wet lab — week one

You receive a report. Not a login.

PHYSIM is not a platform you log into. I run the computational engine on your behalf and deliver a comprehensive, pre-clinical grade report documenting every physical evaluation — ready for your wet lab team, your investors, or your regulatory filing.

Deep Computational Engagement

Bring me your target.

One disease. Weeks of intensive computational evaluation — deeper than anything available in the field. From molecular survival through chronic safety, disease engagement, and patient-level staging. Every gate documented. Every measurement auditable. A report that goes far beyond pharmacokinetics into territory no one else can reach.

Engagement
2–3 Weeks
Intensive computation
7-Gate Gauntlet Report
  • Pharmacokinetic survival (Gate 1)
  • Tissue & organelle penetration (Gate 2-3)
  • Chronic dosage & duration toxicity (Gate 4)
  • Hepatotoxicity proximity scoring (Gate 5)
  • Disease cascade engagement map (Gate 6)
  • Patient stage matching & classification (Gate 7)
Beyond Standard Deliverables
  • Full disease architecture map
  • Chronic toxicity timeline (dosage × duration)
  • DILI proximity score
  • Disease network engagement analysis
  • Drug-stage classification (Repair / Disrupt / Rebuild / Rewire)
  • Resistance window projections
  • Deterministic reproducibility guarantee

All engagements are project-scoped under mutual confidentiality. Serious candidates will receive a specimen report during intake. No subscription. No software license. You receive the report.

Start a Conversation
Exclusive Access

This platform is not for sale.

PHYSIM is not a SaaS product. It is not a dataset you can license. It is not an API you can call. It is a proprietary computational engine — operated exclusively by its architect.

If you are a pharmaceutical company, biotech firm, or research institution with drug candidates that keep failing — or diseases that no one has been able to crack — bring me your toughest problem. I will run the engine. You will receive validated, pre-clinical grade outputs ready for your wet lab.

Bring Me Your Target About the Architect

Years ago, I set out to look at medicine through a different lens.

I knew that human biology didn't operate in the silos that modern medicine uses to categorize it. But to see the connections, I had to understand the systems. So I leveraged the full capabilities of AI — not to replace human insight, but to build scaffolding that could teach me the language of domains whose terminology was foreign but whose concepts were entirely connected.

Through that process, he built cross-domain computational fluency — systematically connecting the biological systems that modern medicine studies in isolation.

PRESTON MCCAULEY

Founder · Architect · Computational Systems Engineer

🧠
Neuroscience & Cognitive Science
Brain waves, TBI, neurodegeneration
🎯
Tumor Biology & Oncology
Molecular oncology, evasion circuits
⚗️
Pharmacology & Comp. Chemistry
SMILES, drug design, ADMET modeling
🧬
Genetics & Epigenetics
Mutation mapping, chromatin remodeling
🔬
Immunology & Virology
Autoimmune pathways, immune privilege
🤖
AI & Systems Engineering
LLM orchestration, agentic architecture
The Methodology

"Before anything ships, I try to break it. That has always been my methodology. Apply rigorous pressure until something fails. Fix it. Break it again."

90
Diseases Mapped
51
Cancer Types
6,675
MCC Confirmed

Meet the Architect

Preston McCauley didn't start in pharma. He started by asking why the statistics across all of medicine weren't good enough — and refusing to accept the answer.

He began with neuroscience — spending years researching brain waves, cognitive science, and traumatic brain injuries. But he pushed further. Each domain he entered deliberately. Each answer revealed the next domain that needed to be understood. Not because it was his field. Because it impacts people he cares about.

That journey — from neuroscience to tumor biology to genetics to pharmacology to immunology — is what made PHYSIM possible. He didn't assemble off-the-shelf ML models. He spent Sundays at the local coffee shop learning SMILES notation, understanding drug metabolism, and connecting the dots between silos of knowledge that the medical establishment has never connected.

🏆
Industry Recognition

In 2025, Preston was nominated as one of the 75 AI Innovators in Dallas — recognized not for following trends, but for building something the industry hadn't seen before: a cross-domain computational engine designed from first principles by someone who refused to stay in one lane.

Not Just Another Drug Platform

Let's be clear: PHYSIM is not "another drug creation platform." There are plenty of those, and their molecules still fail in Phase II. Preston built a synthetic human — a computational pharmacokinetic engine that computes exactly how and where drugs fail inside the body, from gastric degradation to molecular bounce-offs at the Blood-Brain Barrier.

To prove it works, he ran a 100-compound blind gauntlet test. The system achieved 83.9% concordance and 87.5% sensitivity — exceeding the FDA PBPK pilot guidance threshold (≥85%). It correctly computed the hERG toxicity of Terfenadine and the viability of FDA-approved Empagliflozin. The one red X — Vancomycin at the BBB — was intentional. A system that hides its failures isn't a validation. It's a demo.

Computational Finding: Autoimmune Hepatitis Execution Layer

When PHYSIM mapped Autoimmune Hepatitis, the Stage 4 analysis identified an epigenetic execution layer involving gene programs typically associated with ocular immune privilege architecture — not immune genes, not liver genes.

The implication: the disease may be reactivating a deeply conserved immune-silencing mechanism outside its expected tissue context, permanently disabling the liver's regulatory T cells. This is the class of cross-domain structural relationship the engine is designed to surface.

The Bottom Line

Preston holds the Stage 3 and Stage 4 coordinates for 6,675 distinct diseases across multi-billion dollar therapeutic sectors. He is not building a decade-long SaaS startup. He built a proprietary mapping engine and a synthetic human testing environment — and he is offering a capability.

The map exists. The testing environment is built. Let's talk.

Bring Me Your Target

What this engine is built to surface

These findings span oncology, neurodegeneration, autoimmune disease, metabolic disease, and cross-disease convergence analysis. They illustrate the range of computational outputs the engine produces — not isolated curiosities, but repeatable finding classes.

TP53 Upstream Failure

The guardian of the genome doesn't always fail because it is attacked directly. My system surfaced CRY2 — a circadian clock repressor — as a key upstream signal before the literature review confirmed the cascade. When the clock machinery breaks, c-MYC accumulates unchecked and suppresses TP53 output downstream. The visible failure is not the beginning of the problem.

Literature confirmed: Huber et al., Mol Cell 2016 · Chan et al., PNAS 2021
Diabetes — 76/76 Convergence

Evaluated 76 distinct forms of diabetes — Type 1, Type 2, MODY, neonatal, microvascular, syndromic. Every genetic origin. Every clinical presentation. All 76 follow the same 4-stage collapse sequence and terminate at the same 8 nodes. Medicine has treated Type 1 and Type 2 as fundamentally different diseases. The data tells a different story.

Novel finding: GCK→IL2RA bridge — T1/T2 boundary collapse · Not in published literature
Hidden Circuit of Cancer

Mapped a candidate execution layer where cancer sustains division regardless of intervention. Validated the structure six independent ways. Confirmed consistent patterns across 51 cancer types simultaneously. Identified a single convergence point where the circuit may be interrupted. The same evasion architecture. Every cancer. Every time.

Cross-validated across autoimmune, infectious, neurological, and metabolic disease
Duchenne — Ascending Neurodegeneration

One in five thousand boys. Progressive. Fatal. My analysis found Duchenne may be an ascending neurodegenerative disease — not a localized muscle tear. Stage 3 shares the exact failure point as Myasthenia Gravis. Stage 4 shares the terminal pathway with ALS. Four independent disease sweeps confirmed. Three existing therapeutic pipelines. Three different medical worlds. One underlying mechanism.

Cross-validated: MG, Becker MD, ALS, Malignant Hyperthermia
Autoimmune Hepatitis — The Eye Gene Discovery

Mapped a 7-stage cascade where the immune system turns against its own liver. At Stage 4, my analysis identified the epigenetic executors — genes that build the human eye — permanently silencing FOXP3 and eliminating regulatory T cells. Why is a liver disease using eye genes to execute a death sequence? Because the eye carries evolution's master immune off-switch. Certain diseases appear to have learned to steal it.

Most complete computational map of AIH produced outside a wet lab
The Dual Shape of Huntington's

Generated two biological maps of the same gene (HTT). They look completely different. In science, when the same thing has two different shapes, something important is being missed. Every trial has been built from one picture. The map just got bigger.

Dual-axis structural mapping · Novel insight
ERBB2 Escape Architecture

Mapped the complete resistance landscape around HER2 — the oncogene behind aggressive breast cancers. The engine identified 8 genes that fill the void when HER2 is computationally removed. These aren't similar proteins. They're the exact backup circuits the cancer activates when you block the primary target. Standard drugs fail here. This is why.

p95HER2 resistance addressed computationally
12/12 Blind Validation

Tested 12 computationally identified gene relationships against published scientific literature. Zero prior knowledge. The engine found HGF as an EGFR bypass — the #1 documented resistance mechanism with active Phase III trials. Found RAC1 as essential for KRAS. Found BRIP1 directly bound to BRCA1. 12 out of 12 confirmed. Zero false positives.

Statistical significance: p < 0.005 (permutation test)
BRAF Melanoma Axis

Mapped the full BRAF V600E neighborhood — the mutation behind ~50% of melanomas. The engine revealed concentric functional shells: an inner kinase shell, a middle tumor suppressor shell (PTEN, TP53, RB1), and an outer immunotherapy target shell (PD-1, CTLA-4). Bridge nodes between BRAF and TP53 may represent novel combination targets.

36 targets mapped across 9 structural dimensions

The Universal Architecture of Failure

If 51 cancers all use the exact same evasion circuit — what about everything else? What if cancer, autoimmune disease, infection, neurodegeneration, and metabolic disease all fail the same way? Not similar. The same.

Preston stress-tested this architecture against 6,675 distinct human diseases across autoimmune, cancer, infectious, neurological, metabolic, cardiovascular, psychiatric, genetic, and rare disease categories. The pattern didn't break.

McCauley's Convergence Cascade

Across all 6,675 diseases, two highly specific convergence points emerge:

STAGE 3 — BRAKE DESTRUCTION

Ubiquitin ligases (E3) are hijacked to systematically tag and destroy the body's regulatory machinery. The immune brakes are physically dismantled. Confirmed across all mapped diseases.

STAGE 4 — THE EPIGENETIC LOCK

Chromatin remodeling permanently encodes the corrupted instruction into the genome. This is the point of no return. The disease becomes self-sustaining. Confirmed across all mapped diseases.

This is where the disease wins. And this is not where medicine is looking.

Where I am. Where I'm going.

The computational foundation is complete. The following milestones represent the path from validated computational targets to IND-enabling preclinical data — and ultimately, to the clinic.

Phase 1 — Computational Validation
Complete

Disease architecture mapping across 6,675 conditions. Development and validation of the McCauley Convergence Cascade (MCC) framework. Identification and confirmation of Stage 3 (ubiquitin-mediated brake destruction) and Stage 4 (epigenetic lock-in) convergence points across 6,675 distinct human diseases spanning 9 pathological categories.

6,675 MCC Confirmed 51 Cancer Types Mapped 6,675 Disease Architectures
Phase 2 — Synthetic Human Gauntlet
Complete

Construction and validation of a 15-stage computational pharmacokinetic model simulating the complete ADMET cascade — from oral bioavailability and gastric survival through hepatic first-pass metabolism, plasma protein binding, CYP450 interactions, hERG cardiac liability, blood-brain barrier penetration, and renal clearance. Blind validation against 100 known compounds.

83.9% Concordance 87.5% Sensitivity Exceeds FDA PBPK ≥85%
Phase 3 — Wet Lab Validation & Partnership
Current Phase

Seeking institutional partners to validate computationally identified Stage 3 and Stage 4 molecular targets through in vitro assays and biochemical confirmation. Priority therapeutic areas include oncology (pancreatic, glioblastoma), neurodegeneration (Huntington's, Alzheimer's), and autoimmune conditions with high unmet medical need. All target coordinates are proprietary and will be disclosed under mutual CDA.

Seeking Lab Partners CDA Required UTSW Ecosystem Priority
Phase 4 — IND-Enabling Preclinical Development
Upcoming

Following wet lab confirmation of lead target candidates: dose-response profiling, selectivity panels, and preliminary DMPK studies to establish therapeutic index. Preparation of IND-enabling data packages aligned with FDA PBPK pilot program guidance for computational-first submissions.

DMPK Profiling Selectivity Panels FDA PBPK Alignment
Phase 5 — Clinical Translation & Licensing
Future

Out-licensing of validated preclinical assets to pharmaceutical partners for clinical development. Targets spanning oncology, autoimmune, cardiovascular, neurodegeneration, and metabolic indications — each with computationally validated mechanism-of-action, selectivity data, and synthetic human survival profiles.

Licensing Deals Multi-Indication Portfolio Clinical Development

The computational work is done. The targets are identified. What comes next requires partners who can take these coordinates into the physical world. If you operate in preclinical research, drug development, or biotech investment — this is your window.

Discuss Partnership

Who commissions a Computational Disease Analysis

We produce pre-clinical CDA reports for organizations that need computational answers before entering the lab.

🏥

Rare Disease Foundations

You have research grant budgets and a mission to find treatments. We narrow your drug candidates computationally — so every dollar you spend in the lab is aimed at the highest-probability targets.

"We have $2M in research grants — where should we aim them?"
🧪

Biotech Startups

You have a target but limited runway. We validate your hypothesis computationally before you burn capital in the lab — giving you data that strengthens your next raise.

"Can you prove our target before we spend $3M on wet lab?"
💊

Pharma R&D

Your Phase 2 failed. Your pipeline stalled. We find the pathway you missed and identify existing FDA-approved drugs that may already work against your target — without starting from scratch.

"Our trial failed — is there a pathway we didn't see?"
🎓

Academic Researchers

You need computational validation to strengthen grant applications or back up a novel hypothesis. We produce deterministic, citable findings that complement your wet lab data.

"I need computational evidence to strengthen my NIH application"

Questions about the engine

Common questions from pharma partners, biotech investors, and research institutions.

No. Probabilistic models (like standard bioinformatics AI) guess connections based on how frequently terms appear near each other in medical journals. PHYSIM is a deterministic thermodynamic engine. It does not guess. It evaluates the raw structural and physical affinity between biological systems. When it bridges a disease phenotype to a downstream target, it is mapping rigid physical realities—not textual probabilities. This allows the engine to autonomously construct entirely novel disease cascades that have never been written about in any database or publication.
No. PHYSIM is a proprietary computational engine that I operate on your behalf. You bring me a disease target or therapeutic area — I run the full pipeline and deliver validated candidates, pharmacokinetic reports, and ADMET simulation data directly to you. You receive the output, not platform access.
You receive pre-clinical grade reports. For Gauntlet Validation (Tier 1): full 15-stage Elevated ADMET simulation results with quantitative values for each biological checkpoint, pharmacokinetic profiles (bioavailability, half-life, volume of distribution, plasma protein binding, CYP450 interactions, hERG cardiac liability), FDA PBPK alignment scoring, route of administration comparison (oral vs. IV), and failure mode analysis with margin notes. For Candidate Generation (Tier 2+): all of the above, plus a novel molecular candidate designed against your target with structural justification. For Disease Mapping (Tier 3+): biological failure architecture mapping with MCC convergence cross-reference. Everything you need to walk into a wet lab with confidence.
A standard full-campaign engagement — from disease target intake to validated candidate delivery — is measured in weeks, not months or years. The exact timeline depends on disease complexity and the number of targets in scope. This replaces what traditionally takes 4–6 years of pre-clinical work.
The McCauley Convergence Cascade (MCC) is the proprietary computational framework at the heart of PHYSIM. It identifies where chronic diseases converge at the biological level — revealing shared upstream control points across 6,675 diseases and 9 pathological categories. This cross-disease architecture is what allows the engine to find intervention targets the standard literature may miss.
All engagements are conducted under mutual confidentiality agreement. The candidates and reports delivered to you are yours. The engine, methodology, and computational IP remain proprietary. IP terms, licensing, and co-development arrangements are discussed at intake and formalized before any work begins.
PHYSIM is designed for pharmaceutical executives, biotech investors, research institutions, and clinical-stage companies looking to accelerate their pre-clinical pipeline. If you have a disease target that has been expensive or slow to pursue through traditional screening, this engine was built for your problem.
Yes. In addition to designing novel candidates, the engine can evaluate existing FDA-approved compounds against new disease indications. If you have an approved drug and suspect it may work for a different target, PHYSIM will run it through the full 15-stage Gauntlet in the context of the new indication — delivering complete pharmacokinetic, toxicity, and route-of-administration data for the repurposed application. This is particularly valuable for rare disease foundations and biotech programs exploring new uses for known compounds with established safety profiles.
Qualified prospects will receive a specimen report during the intake process. This is a redacted example of an actual Tier 1 Gauntlet Validation — showing the full 15-stage simulation output, pharmacokinetic profiles, route comparison, failure analysis, and calibration anchor validation. It demonstrates exactly what you receive as a deliverable, with compound identity redacted.
No. There are no learned weights, no neural networks, no training data, and no probability distributions. The engine computes physical properties directly from molecular structure using chemical mathematics, then simulates whether the molecule physically survives each biological zone in the human body. Run it 10,000 times on the same molecule — you get the same answer every time. Zero variance. This is physics, not machine learning.
Predictive tools train on historical outcomes and then guess what will happen to a new molecule. They give you a probability — "This molecule has a 73% chance of hERG toxicity." PHYSIM computes the physical properties of the molecule and simulates whether it survives each biological boundary. The analogy is a suspension bridge: you don't predict that removing a load-bearing cable will collapse the bridge. You compute the tension forces, and the math proves the geometry cannot hold. That's not a prediction — it's structural physics. PHYSIM operates the same way on molecules.
It means that across 219,090 individual physical evaluations, the engine's computed outcomes agree with observed clinical reality 97.8% of the time. I deliberately use the word concordance, not accuracy — because the engine is not making predictions that can be "right or wrong." It is computing physics. The 2.2% gap represents the frontier of the approximation — where simplified computational physics hasn't yet fully captured a complex biological phenomenon. That's an engineering challenge, not a prediction error.
GET IN TOUCH

What is your toughest drug problem?

Bring me your hardest target. Fill out the form below and I'll respond within 24 hours.

ELEVATED ADMET

97.8% concordance · Proprietary 15-stage biological computation · pre-wet lab deterministic physics at a resolution the industry has never operated at. Your targets, my engine.