v2026.2.0
Reasoning Summaries connect multi-omic evidence, disease context, and scientific rationale behind every ranked result.
🚀 Introducing Reasoning Summaries
We’re excited to introduce Reasoning Summaries, a new way to explore and understand an object's ranking within your graph.
Reasoning Summaries are powered by Reasoning Kernels, which rank and prioritize data points according to a team’s scientific frameworks and proprietary data. Reasoning Kernels encode how your organization reasons about biology by combining evidence weighting, data selection, and domain-specific assumptions into a reproducible prioritization strategy.
Reasoning Summaries provide a comprehensive, multi-omic view of each object in your graph, directly connected to the Reasoning Kernel used to rank and prioritize it. With a single click on an object within a report, users can now move seamlessly from high-level prioritization to detailed scientific context.
What’s New
🔍 Object-Level Drill-Down from Reports
Select any object, such as a drug, gene, disease, or pathway, within a report to instantly access its full summary without context switching.
📊 Ranking Transparency
Clearly see:
Where the object ranks within the report
How it compares to peer objects
Evidence contributing to its overall score
🧬 Comprehensive Multi-Omic Summary
Each Reasoning Summary summarizes all relevant data connected to the object in the context of the scientific frameworks encoded in the Reasoning Kernel.
This can include:
Proteomic, transcriptomic, epigenomic, and genetic quantitative data
Third-party consortium data
Cross-modal signals unified into a single, interpretable view
📝 Disease-Contextualized Narrative Summary
Review an AI-generated, text-based summary that explains the object in the context of the selected disease, translating complex graph evidence into a coherent scientific narrative.
🧪 Actionable Recommendations
Identify clear next steps with:
Suggested follow-up experiments
Missing or under-represented data types
Gaps in evidence that limit confidence in the object’s ranking
🔗 Full Evidence Provenance
Inspect every underlying data point supporting the object’s role in the disease:
All contributing nodes and edges from the knowledge graph
Complete traceability from raw data to ranking outcome
Example: Target Prioritization in ALS
A discovery team runs a Target Prioritization Reasoning Kernel for amyotrophic lateral sclerosis.
The report ranks hundreds of candidate targets based on integrated genetic association data, patient-derived expression data, functional genomics screens, and clinical evidence, all weighted according to the team’s scientific framework.
When a scientist selects a top-ranked target from the report, the Reasoning Summary shows:
The target’s exact rank within the ALS prioritization Reasoning Kernel
A narrative explanation of how the target is implicated in motor neuron degeneration
Supporting multi-omic evidence, such as differential expression in ALS patient tissue and genetic risk enrichment
Recommendations to strengthen confidence, such as proposed CRISPR perturbation studies and missing cell-type-specific expression data
Full access to all underlying graph data that contributed to the target’s prioritization
This enables the team to quickly validate why a target surfaced, assess confidence, and decide whether to advance it into experimental follow-up.
Why It Matters
As part of the BioBox Decision Operating System, Reasoning Summaries close the gap between prioritization and action. Teams can now:
Trust rankings by understanding why an object scores the way it does
Quickly assess biological rationale without manual data aggregation
Move faster from model output to experimental decision-making
Last updated