BPE tokenization instead of k-mers;
input sequence size is about 4500 nucleotides (512 BPE tokens) compared to 512 nucleotides of DNABERT
pre-training on T2T + Multispecies vs. GRCh38.p13 human genome assembly.

Source code and data: https://github.com/AIRI-Institute/GENA_LM

Paper: https://academic.oup.com/nar/article/53/2/gkae1310/7954523

Examples

How to load pre-trained model for Masked Language Modeling

python

from transformers import AutoTokenizer, AutoModel

tokenizer = AutoTokenizer.from_pretrained('AIRI-Institute/gena-lm-bert-base-t2t-multi')
model = AutoModel.from_pretrained('AIRI-Institute/gena-lm-bert-base-t2t-multi', trust_remote_code=True)

How to load pre-trained model to fine-tune it on classification task

Get model class from GENA-LM repository:

git clone https://github.com/AIRI-Institute/GENA_LM.git

python

from GENA_LM.src.gena_lm.modeling_bert import BertForSequenceClassification
from transformers import AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained('AIRI-Institute/gena-lm-bert-base-t2t-multi')
model = BertForSequenceClassification.from_pretrained('AIRI-Institute/gena-lm-bert-base-t2t-multi')

or you can just download modeling_bert.py and put it close to your code.

OR you can get model class from HuggingFace AutoModel:

python

from transformers import AutoTokenizer, AutoModel
model = AutoModel.from_pretrained('AIRI-Institute/gena-lm-bert-base-t2t-multi', trust_remote_code=True)
gena_module_name = model.__class__.__module__
print(gena_module_name)
import importlib
# available class names:
# - BertModel, BertForPreTraining, BertForMaskedLM, BertForNextSentencePrediction,
# - BertForSequenceClassification, BertForMultipleChoice, BertForTokenClassification,
# - BertForQuestionAnswering
# check https://huggingface.co/docs/transformers/model_doc/bert
cls = getattr(importlib.import_module(gena_module_name), 'BertForSequenceClassification')
print(cls)
model = cls.from_pretrained('AIRI-Institute/gena-lm-bert-base-t2t-multi', num_labels=2)

Model description

GENA-LM (gena-lm-bert-base-t2t-multi) model is trained in a masked language model (MLM) fashion, following the methods proposed in the BigBird paper by masking 15% of tokens. Model config for gena-lm-bert-base-t2t-multi is similar to the bert-base:

512 Maximum sequence length
12 Layers, 12 Attention heads
768 Hidden size
32k Vocabulary size

We pre-trained gena-lm-bert-base-t2t-multi using the latest T2T human genome assembly (https://www.ncbi.nlm.nih.gov/assembly/GCA_009914755.3/). The data was augmented by sampling mutations from 1000-genome SNPs (gnomAD dataset). We also add multispecies genomes from ENSEMBL release 108. The list of used species is here. Pre-training was performed for 1,925,000 iterations with batch size 256 and sequence length was equal to 512 tokens. We modified Transformer with Pre-Layer normalization, but without the final layer LayerNorm.

Evaluation

For evaluation results, see our paper: https://academic.oup.com/nar/article/53/2/gkae1310/7954523

Citation

bibtex

@article{GENA_LM,
    author = {Fishman, Veniamin and Kuratov, Yuri and Shmelev, Aleksei and Petrov, Maxim and Penzar, Dmitry and Shepelin, Denis and Chekanov, Nikolay and Kardymon, Olga and Burtsev, Mikhail},
    title = {GENA-LM: a family of open-source foundational DNA language models for long sequences},
    journal = {Nucleic Acids Research},
    volume = {53},
    number = {2},
    pages = {gkae1310},
    year = {2025},
    month = {01},
    issn = {0305-1048},
    doi = {10.1093/nar/gkae1310},
    url = {https://doi.org/10.1093/nar/gkae1310},
    eprint = {https://academic.oup.com/nar/article-pdf/53/2/gkae1310/61443229/gkae1310.pdf},
}