Skip to content

🔧 Applied Fixes to neural_module.py

📋 Original Problem Summary

When running neural_module.py, you encountered the error: 

✗ Error processing sequence test_sequence_1: H3K27AC

🔍 Diagnosis

I identified 3 main problems:

1️⃣ Incorrect Outputs

  • Problem: Using output names that don't exist (H3K27AC, H3K4ME3, CTCF)
  • Cause: AlphaGenome groups outputs by category
  • Solution: Updated to use correct outputs

Before: 

--outputs H3K27AC H3K4ME3 H3K27ME3 CTCF

After: 

--outputs CHIP_HISTONE CHIP_TF

2️⃣ Incorrect API Method

  • Problem: Using model.predict() which doesn't exist
  • Cause: API uses specific methods
  • Solution: Updated to model.predict_interval()

3️⃣ Missing Parameter

  • Problem: predict_interval() requires ontology_terms
  • Cause: AlphaGenome needs to know which tissues to analyze
  • Solution: Added default terms (brain, liver, heart)

4️⃣ Sequence Sizes ⚠️ CRITICAL

  • Problem: Example sequences were 600 bp (not supported)
  • Cause: AlphaGenome only accepts specific sizes
  • Solution: Updated example to 2048 bp

✅ Applied Fixes

1. Correct AlphaGenome Outputs

Created check_alphagenome_outputs.py to list available outputs:

✓ 11 outputs found:
  - RNA_SEQ, CAGE, PROCAP
  - ATAC, DNASE
  - CHIP_HISTONE (H3K27AC, H3K4ME3, etc.)
  - CHIP_TF (CTCF, etc.)
  - CONTACT_MAPS
  - SPLICE_JUNCTIONS, SPLICE_SITES, SPLICE_SITE_USAGE

2. Updated Configuration

DEFAULT_CONFIG = {
    'supported_lengths': [2048, 16384, 131072, 524288, 1048576],
    'default_outputs': [
        'RNA_SEQ',
        'CAGE',
        'ATAC',
        'CHIP_HISTONE',  # Histone markers
        'CHIP_TF',       # Transcription factors
    ],
}

3. Improved Validation

Now checks if sequence size is supported:

def validate_sequence(seq):
    supported = [2048, 16384, 131072, 524288, 1048576]
    if len(seq) not in supported:
        return False, f"Size {len(seq):,} bp not supported. Valid: {supported}"
    return True, None

4. Correct API Method

# Before (WRONG):
outputs = self.model.predict(interval=interval, requested_outputs=output_types)

# After (CORRECT):
outputs = self.model.predict_interval(
    interval=interval,
    ontology_terms=['UBERON:0000955', 'UBERON:0002107', 'UBERON:0000948'],
    requested_outputs=output_types
)

5. Improved Error Handling

# Now shows clear messages and full traceback
try:
    output_type = getattr(self.dna_client.OutputType, out)
except AttributeError:
    console.print(f"[yellow]⚠ Output '{out}' not available, skipping...[/yellow]")

6. Updated Example File

# Before: example_sequence.fasta (600 bp) ❌
# After: example_sequence.fasta (2048 bp) ✅

📚 New Files Created

  1. scripts/diagnostics/check_alphagenome_outputs.py - List available outputs
  2. scripts/diagnostics/check_dna_client_methods.py - List API methods
  3. docs/historical/alphagenome/AVAILABLE_OUTPUTS.md - Complete output documentation
  4. docs/historical/alphagenome/SUPPORTED_SIZES.md - Guide to sequence sizes
  5. docs/APPLIED_FIXES.md - This file

🎯 How to Use Now

Basic Usage (RNA-seq and ATAC)

genomics alphagenome analyze -- \
    -i example_sequence.fasta \
    -k YOUR_API_KEY \
    -o results/ \
    --outputs RNA_SEQ ATAC

Complete Epigenetic Analysis

genomics alphagenome analyze -- \
    -i sequence_2kb.fasta \
    -k YOUR_API_KEY \
    -o results/ \
    --outputs ATAC CHIP_HISTONE CHIP_TF

Splicing Analysis

genomics alphagenome analyze -- \
    -i sequence_16kb.fasta \
    -k YOUR_API_KEY \
    -o results/ \
    --outputs RNA_SEQ SPLICE_JUNCTIONS SPLICE_SITES

⚠️ Important Points

Supported Sizes (CRITICAL!)

AlphaGenome only accepts these sizes: - ✅ 2,048 bp (2 KB) - ✅ 16,384 bp (16 KB) - ✅ 131,072 bp (128 KB) - ✅ 524,288 bp (512 KB) - ✅ 1,048,576 bp (1 MB)

Any other size will result in an error!

Available Outputs

Use the correct names: - ❌ H3K27AC → ✅ CHIP_HISTONE - ❌ CTCF → ✅ CHIP_TF - ✅ RNA_SEQ (correct) - ✅ ATAC (correct) - ✅ CAGE (correct)

🧪 Verification Test

To confirm everything is working:

# 1. Check available outputs
python scripts/diagnostics/check_alphagenome_outputs.py

# 2. Test with example
genomics alphagenome analyze -- \
    -i example_sequence.fasta \
    -k YOUR_API_KEY \
    -o test_results/ \
    --outputs RNA_SEQ \
    --no-plots

# Expected result:
# ✓ 1/1 sequences processed successfully

📊 Final Test Result

✓ Output directory: results_final_test
✓ 1 sequence(s) found
✓ test_sequence_2kb: 2,048 bp
✓ Connection established successfully!
✓ Making predictions for test_sequence_2kb (2048 bp)...
✓ Requested outputs: RNA_SEQ, ATAC
✓ Analysis completed!
✓ 1/1 sequences processed successfully

🚀 Status

Item Status
Correct outputs ✅ Fixed
API method ✅ Fixed
Ontology terms ✅ Added
Size validation ✅ Implemented
Example file ✅ Updated (2048 bp)
Documentation ✅ Complete
Tests ✅ Passing

📖 Additional Documentation

  • Outputs: See docs/historical/alphagenome/AVAILABLE_OUTPUTS.md
  • Sizes: See docs/historical/alphagenome/SUPPORTED_SIZES.md
  • Check outputs: python scripts/diagnostics/check_alphagenome_outputs.py
  • Check methods: python scripts/diagnostics/check_dna_client_methods.py

💡 Tips

  1. Always use supported sizes (2kb, 16kb, 128kb, 512kb, 1MB)
  2. Check available outputs before use
  3. Use --no-plots for quick tests
  4. Start with RNA_SEQ for tests (faster)

✅ neural_module.py is now fully functional!

Fixes applied in: October 2025