The Science
Evidence quality is the variable that determines target success
Most target failures trace back to shallow or poorly classified evidence. Avenzo applies a structured evidence taxonomy and scoring model to separate high-confidence targets from superficially supported ones.
Evidence Taxonomy
A four-tier classification of oncology evidence
Not all evidence is equal. Avenzo classifies every piece of evidence ingested into one of four tiers based on study design, reproducibility, and translational proximity. Tier weights are applied during composite scoring.
| Tier | Evidence type | Weight in scoring |
|---|---|---|
| T1 | Validated in multiple independent preclinical models with consistent direction | High |
| T2 | Single preclinical model or replicated cell-line data, clear mechanistic rationale | Moderate-high |
| T3 | Computational prediction or single-source literature with limited replication | Moderate |
| T4 | Indirect association, pathway-level inference, or low-n observational data | Low |
Scoring Model
How a composite score is assembled
A target's final score integrates all collected evidence through a three-step process. Each step applies independent logic so that bias in one dimension does not inflate the composite.
Individual evidence records are weighted by tier and aggregated per evidence dimension. A target supported by many T1 records scores higher than one with numerous T4 records on the same dimension.
Dimension scores for preclinical literature, clinical context alignment, and computational convergence are integrated using a bounded weighting formula. Dimensions that converge reinforce each other; divergent signals reduce confidence.
A confidence calibration pass adjusts for corpus coverage gaps. Targets with sparse overall evidence receive a coverage penalty, preventing high-dimensional scores built on shallow records from appearing over-ranked.
Validation Approach
How the model is tested against known outcomes
Avenzo evaluates scoring performance using a held-out validation set of targets with known downstream outcomes. Targets are scored as if by a team conducting initial prioritization, then compared to what the evidence record at that time could have predicted.
The validation framework focuses on ranking fidelity: does the model order high-outcome targets consistently above low-outcome ones? A separate retrospective analysis tracks ranking stability over time as evidence accumulates.
Get Started
Apply evidence depth to your current target list
Request early access to run your own target list through the Avenzo scoring pipeline and receive a structured evidence report.