CB‑TRI Risk Summary: Guide to Reading and Interpreting

This guide explains how to read and interpret the metrics presented in the CB‑TRI Risk Summary (the fast‑fail summary) report. The reports are generated by deterministic, template‑driven modules; they provide computational measurements. Use this guide to understand each section, the meaning of each metric and how the different pieces fit together.

CB‑TRI Risk Summary (Fast‑fail Summary)

1. Overview and Context

The top portion lists metadata about the project: Project ID, Run ID, Date, Intended use, Target, Indication, Modality and Mechanism of action. These fields identify the compound and context of the run. A Context description states how the project fits into the existing therapeutic landscape. None of these fields convey risk; they orient the reader.

2. CB‑TRI – Overall Risk Band and Score

The Computational Biology Translational Risk Index (CB‑TRI) aggregates risk across channels (target biology, clinical precedent, chemistry, safety, manufacturability and quantum). The report provides:

• Risk score – a normalized value between 0 and 1 (displayed as 0–100) representing the observed computational risk magnitude.
• Risk band – a label (LOW, MODERATE, ELEVATED, HIGH or CRITICAL) summarizing the magnitude.
• Evidence strength – how well the executed modules support interpretation.
• Confidence – calibration or model‑support strength; it does not mean likelihood of program success.

For the sample run, the overall CB‑TRI score is 43.2/100 (Moderate). This indicates a mid‑range computational risk under the current evidence; it is not a go/no‑go decision.

3. Channel Meters

Each risk channel has its own score, band and confidence:

Channel	Score (0–100)	Risk band	Confidence	Notes
Target Biology	41.0	Moderate	73.8 %	Represents biological risk associated with modulating the target.
Clinical Precedent	N/A (9 programs)	Context Available	Medium	Lists number of similar programs; not scored.
Chemistry	75.0	Elevated	—	Reflects developability risk from chemical properties.
Safety	36.7	Moderate	—	Aggregates exposure, toxicity, metabolism and off‑target predictions.
Manufacturability	20.0	Low	—	Synthetic feasibility and cost/complexity of manufacture.
Quantum	N/A	N/A	—	Out‑of‑scope; no data.

Use these scores to see which areas drive the overall risk. Higher scores imply greater observed risk, but only within the computational scope.

4. Risk Surface Overview

This section summarizes coverage and highlights the most concerning signals:

• Critical risk signals – number of signals in the CRITICAL band (0 in the example).
• Elevated risk signals – number of signals in the ELEVATED band (4 in the example).
• Module coverage – executed/in‑scope modules over total modules (25/25).
• Evidence profile – whether evidence is SUPPORTED, LIMITED or UNRESOLVED.
• Observed risk signals – total number of non‑low risk signals (4).

This overview tells you how complete the computation was and whether there are unresolved evidence gaps. A low coverage fraction (<1.0) indicates missing modules; MIXED evidence strength means some predictions have limited confidence.

5. Target Biology Risk Snapshot

A high‑level summary of biological risk is provided. It states the overall band (e.g., Moderate with score 0.410) and confidence (e.g., medium with 0.738). Detailed submodule data appear in the computational safety report, but the snapshot lists the top risk drivers (e.g., knockout phenotypes, substrate pleiotropy, expression window) and their evidence IDs. Use this to see what biological factors matter most.

6. Wet‑Lab Readiness

This section lists any wet‑lab modules and whether they are Enabled or Disabled. A status of Disabled simply means that the wet‑lab policy is not configured; no wet‑lab recommendation is made.

7. Projected Failure Mode Signals

Each card describes a potential failure mode predicted to break the program first. Key elements:

• Risk signal description and band – the type of failure and the band (e.g., Exposure collapse / solubility is High).
• Findings – the underlying evidence (e.g., insoluble logS, predicted exposure metrics).
• P(first failure, conditional) – predicted probability that this failure mode will occur first (e.g., 8.3 %).
• Severity (Impact) – estimated impact if it occurs (0–100 %).
• Evidence strength (Confidence) – how well supported the prediction is (0–100 %).
• Interpretation – narrative summarizing the effect (e.g., poor solubility can derail translation).
• Recommended evidence strengthening – what experiments or analyses will de‑risk the failure mode. Action pills summarize estimated cost, turnaround time and expected cost of resolution (ECR).

Work through each card to understand which failure modes are most pressing and what actions could reduce uncertainty. Lower‑ranked modes often have lower severity or smaller probability of being first.

8. Semantics Legend

At the end of the summary, the legend defines key terms:

• Risk score – normalized magnitude; higher scores mean greater observed risk.
• Risk band – categorical label summarizing risk magnitude.
• Evidence strength – adequacy of supporting data.
• Confidence – calibration strength, not program success probability.
• Evidence status – whether data are SUPPORTED, LIMITED, UNRESOLVED or NO DATA.
• Technical integrity – whether the run produced clean auditable artifacts.
• Wet‑lab policy – pass/hold/fail semantics apply only if user‑defined thresholds exist; otherwise wet‑lab policy is disabled.

Always refer to the legend when interpreting terms; it clarifies that these numbers are measurements, not recommendations.

Using the Guides

1. Start with the CB‑TRI overview to understand the overall risk profile and identify which channels contribute most to risk. Use the semantics legend to interpret scores and bands correctly.
2. Dive into the projected failure mode cards for actionable insights. Prioritise modes with higher P(first failure) and severity and review recommended experiments.
3. Consult the computational safety report for detailed metrics behind each risk channel. Examine ADMET predictions, toxicity profiles, metabolism, solubility and manufacturability to validate or challenge the risk summary.
4. Pay attention to uncertainty and evidence strength. Sections indicating Unknown or Skipped or low evidence strength point to data gaps; design studies to fill them.
5. Cross‑reference evidence IDs when you need to trace the origin of a metric or verify the underlying data.
6. Treat predictions as exploratory. These metrics provide hypotheses for early‑stage drug discovery; they are not clinical recommendations.

By following this structured approach, you can systematically evaluate computational risk outputs, identify key liabilities and design evidence‑strengthening experiments.