⚠️ Sequence Sizes Supported by AlphaGenome

🎯 Important Information

AlphaGenome only accepts sequences with specific sizes:

Size	Bytes	Recommended Use
2,048 bp	2 KB	Small regions, quick tests
16,384 bp	16 KB	Individual genes
131,072 bp	128 KB	Multiple genes, clusters
524,288 bp	512 KB	Large genomic regions
1,048,576 bp	1 MB	Extensive chromosomal regions

❌ What Happens with Other Sizes?

If you try to use a sequence of 600 bp, 5000 bp, or any other unlisted size, you'll receive the error:

ValueError: Sequence length 600 not supported by the model. 
Supported lengths: [2048, 16384, 131072, 524288, 1048576]

✅ How to Fix

Option 1: Pad (Add Bases)

If your sequence is smaller than the minimum size (2048 bp), add N bases:

sequence = "ATCGATCG..."  # 600 bp
padded = sequence + "N" * (2048 - len(sequence))  # Now has 2048 bp

Option 2: Truncate

If your sequence is larger but not a supported size, truncate to the next smaller size:

sequence = "ATCGATCG..." * 1000  # 5000 bp
truncated = sequence[:2048]  # Use first 2048 bp

Option 3: Resize to Next Supported Size

def resize_to_supported(seq):
    supported = [2048, 16384, 131072, 524288, 1048576]

    # Find the nearest supported size
    for size in supported:
        if len(seq) <= size:
            if len(seq) < size:
                # Add Ns
                return seq + "N" * (size - len(seq))
            return seq

    # If larger than 1MB, truncate
    return seq[:1048576]

📝 Practical Examples

Example 1: Small Sequence (600 bp → 2048 bp)

# Create sequence with padding in Python
python3 << 'EOF'
seq = "ATCG" * 150  # 600 bp
padded = seq + "N" * (2048 - len(seq))
with open("seq_2kb.fasta", "w") as f:
    f.write(">my_sequence_2kb\n")
    f.write(padded + "\n")
EOF

# Analyze
genomics alphagenome analyze -- -i seq_2kb.fasta -k API_KEY -o results/

Example 2: Typical Gene (~3000 bp → 16384 bp)

# gene.py
gene_seq = open("BRCA1_partial.fasta").read().strip()  # ~3000 bp

# Resize to 16kb
if len(gene_seq) < 16384:
    gene_seq_resized = gene_seq + "N" * (16384 - len(gene_seq))

with open("BRCA1_16kb.fasta", "w") as f:
    f.write(">BRCA1_padded_16kb\n")
    f.write(gene_seq_resized)

python gene.py
genomics alphagenome analyze -- -i BRCA1_16kb.fasta -k API_KEY -o results/

Example 3: Large Region (~100kb → 131kb)

# Extract 131kb region from genome
samtools faidx genome.fa chr1:1000000-1131072 > region_131kb.fasta

# Analyze
genomics alphagenome analyze -- -i region_131kb.fasta -k API_KEY -o results/

🔧 Helper Script for Automatic Resize

Created a helper script that does the resize automatically:

# resize_fasta.py (create this script)

Usage:

python resize_fasta.py input.fasta output.fasta --target 2048

💡 Recommendations

For Variant Analysis

• Use 2048 bp or 16384 bp
• Center the variant in the sequence

For Gene Analysis

• Small gene (<10kb): use 16384 bp
• Large gene: use 131072 bp or 524288 bp

For Regulatory Region Analysis

• Promoter + nearby enhancers: 16384 bp
• Multiple enhancers: 131072 bp

For TAD Domain Analysis

• Use 524288 bp or 1048576 bp
• Ideal for CONTACT_MAPS

⚠️ Important Notes

1. Padding with N: N bases won't affect predictions of real bases
2. Centering: When padding, consider centering your sequence of interest
3. Performance: Larger sequences take longer to process
4. Cost: Larger sequences may consume more of your API quota

🆘 Troubleshooting

Error: "Sequence length X not supported"

Solution: Use one of the 5 supported sizes listed above

Sequence is exactly 2048 bp but still gives error

Possible cause: Invisible characters (spaces, line breaks) Solution:

seq = seq.replace('\n', '').replace(' ', '').strip()

How to check my sequence size?

# Via command line
grep -v ">" sequence.fasta | tr -d '\n' | wc -c

# Via Python
from Bio import SeqIO
rec = SeqIO.read("sequence.fasta", "fasta")
print(len(rec.seq))

📚 Resources

• Check sizes: python scripts/diagnostics/check_alphagenome_outputs.py
• Documentation: See docs/historical/alphagenome/AVAILABLE_OUTPUTS.md
• Valid example: example_sequence.fasta (2048 bp)

---

Updated: neural_module.py now automatically validates sizes and provides clear messages about which sizes are supported.

Last updated: October 2025