title: "{{CARD_TITLE}} — Integration Guidance Card" status: draft updated: {{DATE}} tags: [integration, {{TAGS}}]

{{CARD_TITLE}}¶

Source: {{AUTHORS}} ({{YEAR}}), {{VENUE}}
Type: {{INTEGRATION_PATTERN_TYPE}}
Best for: {{PRIMARY_USE_CASES}}

Problem It Solves¶

Challenge: [Describe the integration challenge this approach addresses]

Solution: [High-level description of the solution approach]

Why traditional approaches fail: [Key limitations of naive or simpler methods]

Core Mechanics¶

1. [Primary Mechanism Name]¶

[Detailed description of how the integration works]

# Code example showing key implementation pattern

Key insight: [Main technical or conceptual contribution]

2. [Secondary Mechanisms if applicable]¶

[Additional details on variants, extensions, or related techniques]

When to Use¶

✅ Use this approach when: - [Condition 1: data characteristics] - [Condition 2: dataset size requirements] - [Condition 3: interpretability needs] - [Condition 4: computational constraints]

✅ Particularly well-suited for: - [Specific application 1] - [Specific application 2] - [Specific application 3]

When to Defer¶

⚠️ Defer to other methods when: - [Condition where this approach is suboptimal] - [Alternative scenario] - [Computational or data constraints]

⚠️ Consider alternatives: - [Alternative 1]: [When to use instead] - [Alternative 2]: [When to use instead]

Adoption in Our Neuro-Omics Pipeline¶

Current Implementation¶

Per-modality setup: - Genetics: [FM choice, embedding dimensions, preprocessing] - Brain: [FM choice, embedding dimensions, preprocessing] - Fusion: [How modalities are combined]

Workflow:

# Step-by-step commands for implementation

Evaluation metrics: - [Metric 1 with rationale] - [Metric 2 with rationale] - [Statistical test for fusion gain]

Integration with ARPA-H BOM¶

[How this approach fits into the Brain-Omics Model escalation strategy]

[Escalation diagram showing where this fits]

Why [start with / escalate to] this approach: - [Rationale 1] - [Rationale 2] - [Rationale 3]

Caveats and Best Practices¶

⚠️ [Caveat 1 Name]¶

Problem: [Description of what can go wrong]

Solution: [How to avoid or mitigate the issue]

# Code example showing correct vs. incorrect approach

⚠️ [Caveat 2 Name]¶

Problem: [Description]

Solution: [Mitigation]

[Repeat for additional caveats]

Practical Implementation Guide¶

Step 1: [Setup Phase]¶

[Detailed instructions for initial setup]

Component	Configuration	Rationale
[Item 1]	[Config]	[Why]
[Item 2]	[Config]	[Why]

Step 2: [Training/Integration Phase]¶

# Detailed code example for core integration step

Step 3: [Evaluation Phase]¶

# Code for evaluating integration performance

[Repeat for additional steps as needed]

Reference Materials¶

Primary paper: - Paper Title (Authors Year) — see ../../generated/kb_curated/papers-md/{{PAPER_SLUG}}.md

Related integration cards: - {{RELATED_CARD_1}} — Brief description (link once created) - {{RELATED_CARD_2}} — Brief description (link once created)

KB integration guides: - Integration Strategy — Overall fusion approach - Design Patterns — Pattern taxonomy - Multimodal Architectures — Model examples

Analysis recipes: - Reference recipe: ../../integration/analysis_recipes/{{RECIPE}}.md - Optional second recipe: ../../integration/analysis_recipes/{{RECIPE}}.md

Model documentation: - Genetics Models — Gene embedding extraction - Brain Models — Brain embedding extraction - Multimodal Models — Fusion architectures

Next Steps in Our Pipeline¶

[Phase 1] — [Description and deliverable]
[Phase 2] — [Description and deliverable]
[Phase 3] — [Description and deliverable]
[Phase 4] — [Description and deliverable]
[Phase 5] — [Description and deliverable]

Success criteria for escalation: - [Quantitative criterion 1] - [Quantitative criterion 2] - [Qualitative criterion]

Key Takeaways¶

[Takeaway 1] — [Explanation]
[Takeaway 2] — [Explanation]
[Takeaway 3] — [Explanation]
[Takeaway 4] — [Explanation]
[Takeaway 5] — [Explanation]

Bottom line: [One-sentence summary of when and why to use this integration approach]