DISEASE PROGRAM 01 • METABOLIC HEPATOLOGY

Non-Alcoholic Fatty Liver Disease (NAFLD / NASH)

NAFLD prevalence in India stands at 38–40% nationally, surging past 65% in urban cohorts. We evaluated our 5-layer platform on our researched data encompassing validated disease cases and healthy controls. Below are our preliminary findings, explainable diagnostic features, and real-time trial simulation engines.

Launch Patient Simulator Explore Drug Effect Engine

VALIDATED PROOF-OF-CONCEPT

1. Clinical Decision Optimization & Biomarker Attributions

Demonstrating cost-optimal threshold calibration (θ*=0.40), global transaminase explainability, and synthetic patient marginal fidelity on our researched data.

Asymmetric Cost Optimization

Penalizes missed cases 5x over false alarms, identifying θ*=0.40 as the minimum clinical loss operating point.

Biomarker Explainability

Identifies Alkaline Phosphatase, Total Bilirubin, and Alamine Aminotransferase as primary risk drivers.

100% Synthetic Patient Fidelity

Continuous-discrete Bayesian Network DAG preserving authentic transaminase log-log dependencies without memorizing data.

INTERACTIVE TOOL

2. Clinical Decision & Progression Simulator

Adjust serum laboratory biomarkers below to observe real-time ensemble inference, conformal prediction coverage, and 5-year disease progression estimates.

Patient Age Adult 45y

18 yrs85 yrs

Total Bilirubin Elevated 1.8

0.4 mg/dL15.0 mg/dL

Direct Bilirubin Cholestatic 0.7

0.1 mg/dL8.0 mg/dL

Alkaline Phosphatase Elevated 220

50 U/L800 U/L

ALT (SGPT) Cytolysis 58

10 U/L400 U/L

AST (SGOT) Injury 62

10 U/L400 U/L

De Ritis Ratio (AST / ALT) 1.07

Reversal Alert: Ratio > 1.0 indicates sinusoidal bridging fibrosis and architectural distortion.

Serum Albumin Normal 3.4

1.5 g/dL5.5 g/dL

Total Serum Proteins Normal 6.7

3.5 g/dL9.5 g/dL

Clinical Presets:

--% CALCULATING

95% Credible Interval: [--% – --%]

Conformal Set: Calculating...

Loading conformal calibration guarantees...

Local SHAP Feature Attribution Impact on Risk Log-Odds

5-Year Semi-Markov Disease Evolution Year 5 Horizon

Y0Y1Y2Y3Y4Y5

S0: Healthy / Reversible Steatosis --%

S1: Active Steatohepatitis --%

S2: Advanced Bridging Fibrosis --%

S3: End-Stage Cirrhosis / Decompensation --%

THERAPEUTIC PHARMACOMETRICS

3. In-Silico Drug Effect & Biomarker Clearance Simulator

Evaluate candidate therapeutics against Indian NAFLD disease progression. Select a mechanistic drug class or enter custom hazard ratios to simulate 5-year event reduction, control-arm patient cutbacks, and pharmacokinetic clearance kinetics.

Therapeutic Hazard Ratio (HR) 0.65

0.40 (High Efficacy)0.95 (Marginal)

Daily Target Dose 50 mg

10 mg100 mg

Trial Simulation Follow-Up Period 36 Months

12 Months60 Months (5 Years)

THR-β selective agonist targeting hepatic lipotoxicity and steatohepatitis resolution without systemic thyrotoxicosis.

Simulated Relative Risk Reduction

-35.0%

POPULATION CALIBRATED

Standard of Care / Placebo Control (Physical Baseline)

21.0% Decompensation Rate at 5 Years

Active Treatment Arm (In-Silico Simulated Cohort)

13.6% Decompensation Rate

48-Week Longitudinal ALT Clearance Trajectory Comparing Physical Control Arm (Placebo) vs In-Silico Active Treatment Arm

Placebo Active NCE Safe (<35 U/L)

Projected Serum ALT (SGPT)

42 U/L

↓ 72% from baseline

Projected Total Bilirubin

1.3 mg/dL

↓ 69% from baseline

45% Control Cut Reduces physical human control arm enrollment requirement.

36% Fibrosis Reversal Simulated virtual patients achieving histological fibrosis regression.

SCOPE OF DELIVERY

4. Platform Deliverables for NAFLD

The complete technical, mathematical, and clinical package delivered by Silicon Arm for hepatic clinical trial simulation.

1. Cloud-Portable Data Harmonization Engine

Standardized pipeline normalizing raw hospital biochemistry panels to unified clinical data models with automated de-identification adhering to India's DPDP regulations.

2. Cost-Sensitive Diagnostic Risk Classifier

Trained and calibrated stacking ensemble operating at θ*=0.40 to minimize missed clinical cases while maintaining high diagnostic sensitivity.

3. Biomarker Explainability & Audit Tooling

Global SHAP feature attribution suite and patient-level clinical flag reports providing transparent mathematical interpretability for investigators and ethics boards.

4. Synthetic Virtual Cohort Generator

Continuous-discrete Bayesian Network DAG & Gaussian Copula generating virtual patients with 100% Kolmogorov-Smirnov test pass rates, preserving non-linear co-dependencies without memorizing training records.

5. Semi-Markov Virtual Control Arm Engine

4-state Hidden Semi-Markov Model with duration-dependent Weibull hazards simulating 5-year event-free survival under standard-of-care vs. active treatment, coupled with ODE pharmacokinetic clearance.

6. Calibrated Uncertainty & Triage Layer

Conformal Prediction Engine providing 95% nominal test coverage guarantees, paired with automated triage for ambiguous borderline patients ({0, 1}).

7. Institutional Deployment & Retrospective Trial Validation

Containerized package ready for calibration and deployment in partner research laboratories for retrospective clinical study validation.