AlphaGenome MCP Server

Instructions

Implementation Reference

• scripts/alphagenome_bridge.py:297-321 (handler)

  Core handler function implementing predict_tissue_specific tool logic: iterates over tissues, calls predict_variant_effect for each, aggregates tissue-specific impact results.

  def predict_tissue_specific(client, params: Dict[str, Any]) -> Dict[str, Any]:
      """
      Predict variant effects across multiple tissues.
      """
      tissues = params.get('tissues', ['brain', 'liver', 'heart'])
  
      results = {}
      for tissue in tissues:
          tissue_params = params.copy()
          tissue_params['tissue_type'] = tissue
          try:
              result = predict_variant_effect(client, tissue_params)
              results[tissue] = {
                  'expression_impact': result['predictions'].get('rna_seq', {}).get('fold_change', 0),
                  'splice_impact': result['predictions'].get('splice', {}).get('delta', 0),
                  'impact_level': result['interpretation']['impact_level']
              }
          except Exception as e:
              print(f"Warning: Failed to predict for tissue {tissue}: {e}", file=sys.stderr)
              results[tissue] = {'error': str(e)}
  
      return {
          'variant': f"{params.get('chromosome')}:{params.get('position')}{params.get('ref')}>{params.get('alt')}",
          'tissue_results': results
      }

• src/tools.ts:198-244 (schema)

  Tool schema definition including name, description, and inputSchema for validation (chromosome, position, ref, alt, optional tissues).

  export const PREDICT_TISSUE_SPECIFIC_TOOL: Tool = {
    name: 'predict_tissue_specific',
    description: `Predict variant effects across multiple tissues.
  
  Compares regulatory impact in different tissues to identify tissue-specific effects.
  
  Default tissues: brain, liver, heart (customizable)
  
  Returns impact levels and expression changes for each tissue.
  
  Perfect for: understanding tissue-specific disease mechanisms, prioritizing relevant tissues.
  
  Example: "Compare rs429358 effects in brain, liver, and heart"`,
    inputSchema: {
      type: 'object',
      properties: {
        chromosome: {
          type: 'string',
          description: 'Chromosome',
          pattern: '^chr([1-9]|1[0-9]|2[0-2]|X|Y)$',
        },
        position: {
          type: 'number',
          description: 'Genomic position',
          minimum: 1,
        },
        ref: {
          type: 'string',
          description: 'Reference allele',
          pattern: '^[ATGCatgc]+$',
        },
        alt: {
          type: 'string',
          description: 'Alternate allele',
          pattern: '^[ATGCatgc]+$',
        },
        tissues: {
          type: 'array',
          items: {
            type: 'string',
          },
          description: 'List of tissues to test (default: brain, liver, heart)',
        },
      },
      required: ['chromosome', 'position', 'ref', 'alt'],
    },
  };

• src/index.ts:154-158 (registration)

  MCP server handler registration: switch case dispatching to AlphaGenome client predictTissueSpecific method.

  case 'predict_tissue_specific': {
    const result = await getClient().predictTissueSpecific(args);
    return {
      content: [{ type: 'text', text: JSON.stringify(result, null, 2) }],
    };

• src/tools.ts:709-730 (registration)

  Tool registration in ALL_TOOLS array export, including PREDICT_TISSUE_SPECIFIC_TOOL (line 713).

  export const ALL_TOOLS: Tool[] = [
    PREDICT_VARIANT_TOOL,
    BATCH_SCORE_TOOL,
    ASSESS_PATHOGENICITY_TOOL,
    PREDICT_TISSUE_SPECIFIC_TOOL,
    COMPARE_VARIANTS_TOOL,
    PREDICT_SPLICE_IMPACT_TOOL,
    PREDICT_EXPRESSION_IMPACT_TOOL,
    ANALYZE_GWAS_LOCUS_TOOL,
    COMPARE_ALLELES_TOOL,
    BATCH_TISSUE_COMPARISON_TOOL,
    PREDICT_TF_BINDING_IMPACT_TOOL,
    PREDICT_CHROMATIN_IMPACT_TOOL,
    COMPARE_PROTECTIVE_RISK_TOOL,
    BATCH_PATHOGENICITY_FILTER_TOOL,
    COMPARE_VARIANTS_SAME_GENE_TOOL,
    PREDICT_ALLELE_SPECIFIC_EFFECTS_TOOL,
    ANNOTATE_REGULATORY_CONTEXT_TOOL,
    BATCH_MODALITY_SCREEN_TOOL,
    GENERATE_VARIANT_REPORT_TOOL,
    EXPLAIN_VARIANT_IMPACT_TOOL,
  ];

• scripts/alphagenome_bridge.py:806-807 (registration)

  Python bridge dispatch: routes 'predict_tissue_specific' action to the handler function.

  elif action == 'predict_tissue_specific':
      result = predict_tissue_specific(client, params)