1. Environment

1.1 Create a new virtual environment

conda create -n pt36 python=3.6
conda activate pt36

1.2 Install packages

pip install torch==1.10.1+cu111 torchvision==0.11.2+cu111 torchaudio==0.10.1 -f https://download.pytorch.org/whl/cu111/torch_stable.html

cd PLANNER
python -m pip install --editable .
python -m pip install -r requirements.txt

2. Fine-tune and predict

2.1 Download DNABERT

Here are DNABERT models:

DNABERT3

DNABERT4

DNABERT5

DNABERT6

If you want to use DNABERT models, please cite the following publication:

Yanrong Ji, Zhihan Zhou, Han Liu, Ramana V Davuluri, DNABERT: pre-trained Bidirectional Encoder Representations from Transformers model for DNA-language in genome, Bioinformatics, 2021;, btab083, https://doi.org/10.1093/bioinformatics/btab083

2.2 Data processing

If you are going to fine-tune DNABERT with your own data, please process your data into kmer format. The example is shown in below.

sequence	label
TGG GGA GAG AGG GGT	1
CAG AGC GCC CCC CCA	0
ATT TTG TGG GGA GAG	0

2.3 Fine-tune with pre-trained model and predict

We use DNABERT-3 model as example.

python fine-tune.py

3. Ensemble

Python pipeline.py --cell_type A

All model we trained are provided at web server http://planner.unimelb-biotools.cloud.edu.au/

If you want to use PLANNER, please cite the following publication: C. Wang et al., "PLANNER: A Multi-Scale Deep Language Model for the Origins of Replication Site Prediction," in IEEE Journal of Biomedical and Health Informatics, vol. 28, no. 4, pp. 2445-2454, April 2024, doi: 10.1109/JBHI.2024.3349584