Desk screen

Desk Screen Report: mcp-proto-okn

Summary

This manuscript describes mcp-proto-okn, a Python-based Model Context Protocol (MCP) server that enables natural-language access to biomedical and scientific knowledge graphs hosted on the OKN Fabric. The work is presented as a software tool paper with two illustrative case studies.

Scope Assessment

In Scope for In Silico: Yes. The manuscript describes original research software with a clear methodological contribution (MCP-based federation of heterogeneous knowledge graphs) and provides concrete evidence of utility through documented case studies. It makes checkable claims about functionality and demonstrates them with reproducible workflows.

Threshold Issues

1. Completeness and Verifiability

The manuscript references:

• Public GitHub repository (https://github.com/sbl-sdsc/mcp-proto-okn)
• Documented example transcripts with full chat logs
• Public OKN Fabric endpoints and registry
• Reproducible case studies with specific dataset identifiers (e.g., OSD-244)

All central evidence appears inspectable. ✓

2. Fundamental Soundness

The technical approach is sound:

• MCP as a standard interface for LLM-tool integration is well-established
• The routing, schema inspection, and ontology expansion strategies are reasonable
• The use of UberGraph for descendant expansion is appropriate and well-motivated
• Multi-graph coordination via SPARQL federation is standard practice

No fundamental design flaws are evident. ✓

3. Clarity and Presentation

The manuscript is clearly written, well-structured, and honest about limitations (endpoint availability, schema heterogeneity, identifier coverage). The workflow diagram and case studies effectively communicate the contribution. ✓

4. Novelty and Contribution

The work is incremental but legitimate:

• MCP-based access to knowledge graphs is not entirely novel (TogoMCP is cited as prior work)
• However, the application to the fragmented Proto-OKN ecosystem with 30+ heterogeneous graphs, automatic ontology expansion, and multi-graph coordination represents a meaningful extension
• The case studies demonstrate practical utility beyond what prior work (TogoMCP, Emonet et al.) addressed

The contribution is appropriately scoped and acknowledged. ✓

5. Evidence-Claims Alignment

• Claims about functionality are supported by working code and documented examples
• Claims about utility are supported by two detailed case studies with real data
• Limitations are stated (endpoint availability, schema variation)
• No overclaiming is evident

✓

Minor Observations (Not Grounds for Desk Rejection)

1. Evaluation scope: The manuscript presents two case studies rather than systematic benchmarking. This is acceptable for a tool paper, but full review should assess whether the evidence is sufficient to support the utility claims.

2. Comparison to related work: The distinction from TogoMCP could be slightly more detailed (e.g., quantitative comparison of graph coverage, schema heterogeneity handled). This is a minor point for reviewers to consider.

3. Reproducibility details: The manuscript could specify Python version, FastMCP version, and dependency versions more explicitly, though the GitHub repository likely contains this.

Venue Fit

This manuscript is well-suited to In Silico:

• It is original research software with a clear methodological contribution
• All central claims can be evaluated from the manuscript and public materials
• It is neither a clinical trial, diagnostic guidance, nor marketing material
• The work is sound enough to merit expert review, even if some claims require scrutiny

---

DESK DECISION: proceed

Rationale: This is a legitimate software contribution with adequate evidence of functionality and utility. While the novelty is incremental (building on prior MCP-based approaches), the application to the heterogeneous Proto-OKN ecosystem and the specific technical solutions (ontology expansion, multi-graph coordination) represent a meaningful advance. The manuscript is complete, clearly presented, and all central claims are verifiable. Send to full review.