--- license: apache-2.0 tasks: - genomic-sequence-modeling frameworks: - jax language: - en - zh tags: - OneScience - Life Sciences - Genomics - DNA Sequence Model - Variant Effect Prediction - AlphaGenome datasets: - OneScience/alphagenome_dataset ---

AlphaGenome

# Model Introduction AlphaGenome is a DNA sequence model proposed by Google DeepMind. It can take DNA intervals up to 1 Mbp as input and predict multiple classes of genomic functional signals for track prediction and regulatory variant effect scoring. Paper: Advancing regulatory variant effect prediction with AlphaGenome https://www.nature.com/articles/s41586-025-10014-0 # Model Description AlphaGenome is implemented based on JAX / Flax and supports genomic interval inference, variant effect scoring, track evaluation, and fine-tuning examples. This model package is accompanied by the ModelScope dataset `OneScience/alphagenome_dataset`, which can be used for quick local verification. # Applicable Scenarios | Scenario | Description | | :---: | :--- | | Genomic interval prediction | Input a reference genome FASTA, chromosome, and interval coordinates, and output predicted tracks for ATAC, DNase, CAGE, RNA-seq, ChIP, and other signals | | Variant effect scoring | Input a VCF or built-in example variants, compare prediction differences between reference and variant sequences, and generate a variant scoring table | | Track prediction evaluation | Use validation data from the AlphaGenome dataset to calculate regression evaluation metrics for different assay bundles | | Fine-tuning experiments | Use a custom reference genome, interval CSV, and BigWig signal files to verify the fine-tuning workflow | | ModelScope / OneCode runtime | After downloading the model project and accompanying dataset, quickly verify script connectivity in a biology-domain runtime environment | # Usage Instructions ## 1. OneCode Usage You can experience intelligent one-click AI4S programming through the OneCode online environment: [Click to experience intelligent one-click AI4S programming](https://web-2069360198568017922-iaaj.ksai.scnet.cn:58043/home) ## 2. Manual Installation and Usage **Hardware Requirements** - GPU or DCU runtime is recommended. - CPU can be used for import checks and small-configuration connectivity verification, but full training and inference are slow. - DCU users need to install DTK in advance. DTK 25.04.2 or later is recommended, or the OneScience-recommended version matching the current cluster. **Environment Check** - NVIDIA GPU: ```bash nvidia-smi ``` - Hygon DCU: ```bash hy-smi ``` ### Download the Model Package ```bash modelscope download --model OneScience/alphagenome --local_dir ./alphagenome cd alphagenome ``` ### Install the Runtime Environment **DCU Environment** ```bash # First activate DTK and Conda conda create -n onescience311 python=3.11 -y conda activate onescience311 # Supports uv installation pip install onescience[bio-dcu] -i http://mirrors.onescience.ai:3141/pypi/simple/ --trusted-host mirrors.onescience.ai ``` After installation, return to the model package directory: ```bash cd ./alphagenome ``` ### Training and Inference Data Introduction The OneScience community has uploaded the data required for AlphaGenome inference, evaluation, and fine-tuning to ModelScope: [OneScience/alphagenome_dataset](https://modelscope.cn/datasets/OneScience/alphagenome_dataset). After downloading, place the data in the `data/` directory of the model package. ```bash modelscope download --dataset OneScience/alphagenome_dataset --local_dir ./data ``` ### Training Weights The repository includes `weight/alphagenome-all-folds`, and all scripts also support specifying model weights through `--model_dir`. ### Prepare Weights If using local weights, place the AlphaGenome Orbax checkpoint in the following directory: ```text weight/ alphagenome-all-folds/ _CHECKPOINT_METADATA _METADATA ... ``` If running in a shared runtime environment, you can also reuse a unified directory through environment variables: ```bash export ONESCIENCE_MODELS_DIR=/path/to/onescience/models export ONESCIENCE_DATASETS_DIR=/path/to/onescience/datasets ``` The scripts will preferentially read: - `${ONESCIENCE_MODELS_DIR}/AlphaGenome/alphagenome-all-folds` - `${ONESCIENCE_DATASETS_DIR}/AlphaGenome` If the above environment variables are not set, the defaults under the current model package are read: - `weight/alphagenome-all-folds` - `data/` ### Interval Inference ```bash bash scripts/inference.sh ``` Equivalent Python command example: ```bash python scripts/run_inference.py \ --fasta_path ./data/reference/HOMO_SAPIENS/GRCh38.p13.genome.fa \ --model_dir ./weight/alphagenome-all-folds \ --chromosome chr19 \ --start 10587331 \ --end 11635907 \ --output_dir ./outputs ``` Inference results will be saved to `outputs/`. ### Variant Effect Scoring ```bash bash scripts/run_variant.sh ``` When specifying VCF input, you can use: ```bash python scripts/run_variant_scoring.py \ --vcf_path ./data/example.vcf \ --fasta_path ./data/reference/HOMO_SAPIENS/GRCh38.p13.genome.fa \ --model_dir ./weight/alphagenome-all-folds \ --output_dir ./outputs_variant ``` Scoring results will be saved as CSV files. ### Track Prediction Evaluation ```bash bash scripts/run_track.sh ``` You can also explicitly specify the data and output paths: ```bash python scripts/run_track_prediction_eval.py \ --model_dir ./weight/alphagenome-all-folds \ --model_version ALL_FOLDS \ --data_dir ./data/v1/train \ --output_path ./outputs_track/eval_results.csv ``` ### Fine-tuning Example ```bash python scripts/run_finetuning.py \ --fasta_path ./data/reference/HOMO_SAPIENS/GRCh38.p13.genome.fa \ --regions_csv ./data/finetune_regions.csv \ --bigwig_paths ./data/sample_atac.bw \ --output_dir ./finetuned_model \ --num_steps 1000 \ --batch_size 2 ``` # Data Format The ModelScope dataset `OneScience/alphagenome_dataset` is recommended to be downloaded to `data/` under the model package. The default structure is as follows: ```text data/ reference/ HOMO_SAPIENS/ GRCh38.p13.genome.fa GRCh38.p13.genome.fa.fai v1/ train/ ... ``` Where: - `reference/HOMO_SAPIENS/GRCh38.p13.genome.fa` is the human reference genome FASTA. - `.fai` is the FASTA index file. - `v1/train/` is the data directory used for track prediction evaluation. - Custom fine-tuning also requires preparing an interval CSV file with column names `chromosome,start,end`, as well as one or more BigWig signal files. # Official OneScience Information | Platform | OneScience Main Repository | Skills Repository | | --- | --- | --- | | Gitee | https://gitee.com/onescience-ai/onescience | https://gitee.com/onescience-ai/oneskills | | GitHub | https://github.com/onescience-ai/OneScience | https://github.com/onescience-ai/oneskills | # Citations and License - This repository is based on the AlphaGenome open-source model and provides DCU adaptation. The related source code uses the Apache License 2.0. - For scientific research, cite the original paper: [Advancing regulatory variant effect prediction with AlphaGenome](https://www.nature.com/articles/s41586-025-10014-0).