Skip to content

DNALLM Documentation

Mutagenesis

zhangtaolab/DNALLM

In-Silico Mutagenesis¶

This tutorial demonstrates saturation mutagenesis analysis: systematically mutating each position in a DNA sequence to measure the effect on model predictions. This reveals critical nucleotides and functional regions.

Full Notebook¶

View Full Notebook

Prerequisites¶

uv pip install -e '.[base,inference,cuda124]'

Load Configuration and Model¶

from dnallm import load_config, load_model_and_tokenizer

configs = load_config("./inference_config.yaml")

model_name = "zhangtaolab/plant-dnagpt-BPE-promoter_strength_protoplast"
model, tokenizer = load_model_and_tokenizer(
    model_name,
    task_config=configs['task'],
    source="modelscope"
)

Initialize Mutagenesis Analyzer¶

from dnallm import Mutagenesis

mutagenesis = Mutagenesis(config=configs, model=model, tokenizer=tokenizer)

Generate Saturation Mutagenesis Library¶

Replace each position with all four nucleotides:

sequence = (
    "AATATATTTAATCGGTGTATAATTTCTGTGAAGATCCTCGATACTTCATATAAGAGATTTTGAGAGAGAGAGAGA"
    "ACCAATTTTCGAATGGGTGAGTTGGCAAAGTATTCACTTTTCAGAACATAATTGGGAAACTAGTCACTTTACTAT"
    "TCAAAATTTGCAAAGTAGTC"
)

mutagenesis.mutate_sequence(sequence, replace_mut=True)
print(mutagenesis.sequences)

Evaluate Mutations¶

Compute predictions for all mutated sequences:

preds = mutagenesis.evaluate(strategy="mean")

Visualize Mutation Effects¶

pmut = mutagenesis.plot(preds, save_path="plot_mut_effects.pdf")

The heatmap shows how each nucleotide substitution at each position affects the predicted score. Bright spots indicate critical positions where mutations cause large changes.