multimolecule/bpnet

1---2datasets:3- multimolecule/bpnet-oskn4library_name: multimolecule5license: agpl-3.06pipeline: regulatory-profile7pipeline_tag: other8tags:9- Biology10- DNA11- dna12widget:13- example_title: tumor protein p5314  pipeline_tag: regulatory-profile15  sequence_type: DNA16  task: regulatory-profile17  text: ACTCCCCTGCCCTCAACAAGATGTTTTGCCAACTGGCCAAGACCTGCCCTGTGCAGCTGTGGGTTGATTCCACACCCCCGCCCGGCACCCGCGTCCGCGCCATGGCCATCTACAAGCAGTCACAGCACATGACGGAGGTTGTGAGGCGCTGCCCCCACCATGAGCGCTGCTCAGATAGCGATGG18- example_title: BRCA1 DNA repair associated19  pipeline_tag: regulatory-profile20  sequence_type: DNA21  task: regulatory-profile22  text: TCATTGGAACAGAAAGAAATGGATTTATCTGCTCTTCGCGTTGAAGAAGTACAAAATGTCATTAATGCTATGCAGAAAATCTTAGAGTGTCCCATCTGG23- example_title: hemoglobin subunit beta24  pipeline_tag: regulatory-profile25  sequence_type: DNA26  task: regulatory-profile27  text: CATTTGCTTCTGACACAACTGTGTTCACTAGCAACCTCAAACAGACACCATGGTGCATCTGACTCCTGAGGAGAAGTCTGCCGTTACTGCCCTGTGGGGCAAGGTGAACGTGGATGAAGTTGGTGGTGAGGCCCTGGGCAGG28- example_title: CF transmembrane conductance regulator29  pipeline_tag: regulatory-profile30  sequence_type: DNA31  task: regulatory-profile32  text: ACTTCACTTCTAATGGTGATTATGGGAGAACTGGAGCCTTCAGAGGGTAAAATTAAGCACAGTGGAAGAATTTCATTCTGTTCTCAGTTTTCCTGGATTATGCCTGGCACCATTAAAGAAAATATCATCTTTGGTGTTTCCTATGATGAATATAGATACAGAAGCGTCATCAAAGCATGCCAACTAGAAGAG33- example_title: telomerase reverse transcriptase34  pipeline_tag: regulatory-profile35  sequence_type: DNA36  task: regulatory-profile37  text: CGCGGGGGTGGCCGGGGCCAGGGCTTCCCACGTGCGCAGCAGGACGCAGCGCTGCCTGAAACTCGCGCCGCGAGGAGAGGGCGGGGCCGCGGAAAGGAAGGGGAGGGGCTGGGAGGGCCCGGAGGGGGCTGGGCCGGGGACCCGGGAGGGGTCGGGACGGGGCGGGGTCCGCGCGGAGGAGGCGGAGCTGGAAGGTGAAGGGGCAGGACGGGTGCCCGGGTCCCCAGTCCCTCCGCCACGTGGGAAGCGCGGTCCTGGGCGTCTGTGCCCGCGAATCCACTGGGAGCCCGGCCTGGCCCCGACAGCGCAGCTGCTCCGGGCGGACCCGGGG38- example_title: KRAS proto-oncogene39  pipeline_tag: regulatory-profile40  sequence_type: DNA41  task: regulatory-profile42  text: GCCTGCTGAAAATGACTGAATATAAACTTGTGGTAGTTGGAGCTGGTGGCGTAGGCAAGAGTGCCTTGACGATACAGCTAATTCAGAATCATTTTGTGGACGAATATGATCCAACAATAGAG43- example_title: prion protein (Kanno blood group)44  pipeline_tag: regulatory-profile45  sequence_type: cDNA46  task: regulatory-profile47  text: ATGGCGAACCTTGGCTGCTGGATGCTGGTTCTCTTTGTGGCCACATGGAGTGACCTGGGCCTCTGC48- example_title: interleukin 1049  pipeline_tag: regulatory-profile50  sequence_type: cDNA51  task: regulatory-profile52  text: ATGCACAGCTCAGCACTGCTCTGTTGCCTGGTCCTCCTGACTGGGGTGAGGGCC53- example_title: Zaire ebolavirus54  pipeline_tag: regulatory-profile55  sequence_type: cDNA56  task: regulatory-profile57  text: AATGTTCAAACACTTTGTGAAGCTCTGTTAGCTGATGGTCTTGCTAAAGCATTTCCTAGCAATATGATGGTAGTCACAGAGCGTGAGCAAAAAGAAAGCTTATTGCATCAAGCATCATGGCACCACACAAGTGATGATTTTGGTGAGCATGCCACAGTTAGAGGGAGTAGCTTTGTAACTGATTTAGAGAAATACAATCTTGCATTTAGATATGAGTTTACAGCACCTTTTATAGAATATTGTAACCGTTGCTATGGTGTTAAGAATGTTTTTAATTGGATGCATTATACAATCCCACAGTGTTAT58- example_title: SARS coronavirus59  pipeline_tag: regulatory-profile60  sequence_type: cDNA61  task: regulatory-profile62  text: ATGTTTATTTTCTTATTATTTCTTACTCTCACTAGTGGTAGTGACCTTGACCGGTGCACCACTTTTGATGATGTTCAAGCTCCTAATTACACTCAACATACTTCATCTATGAGGGGGGTTTACTATCCTGATGAAATTTTTAGATCAGACACTCTTTATTTAACTCAGGATTTATTTCTTCCATTTTATTCTAATGTTACAGGGTTTCATACTATTAATCATACGTTTGACAACCCTGTCATACCTTTTAAGGATGGTATTTATTTTGCTGCCACAGAGAAATCAAATGTTGTCCGTGGTTGGGTTTTTGGTTCTACCATGAACAACAAGTCACAGTCGGTGATTATTATTAACAATTCTACTAATGTTGTTATACGAGCATGTAACTTTGAATTGTGTGACAACCCTTTCTTTGCTGTTTCTAAACCCATGGGTACACAGACACATACTATGATATTCGATAATGCATTTAAATGCACTTTCGAGTACATATCT63- example_title: insulin64  pipeline_tag: regulatory-profile65  sequence_type: cDNA66  task: regulatory-profile67  text: ATGGCCCTGTGGATGCGCCTCCTGCCCCTGCTGGCGCTGCTGGCCCTCTGGGGACCTGACCCAGCCGCAGCCTTTGTGAACCAACACCTGTGCGGCTCACACCTGGTGGAAGCTCTCTACCTAGTGTGCGGGGAACGAGGCTTCTTCTACACACCCAAGACCCGCCGGGAGGCAGAGGACCTGCAGGTGGGGCAGGTGGAGCTGGGCGGGGGCCCTGGTGCAGGCAGCCTGCAGCCCTTGGCCCTGGAGGGGTCCCTGCAGAAGCGTGGCATTGTGGAACAATGCTGTACCAGCATCTGCTCCCTCTACCAGCTGGAGAACTACTGCAACTAG68- example_title: cyclin dependent kinase inhibitor 2A69  pipeline_tag: regulatory-profile70  sequence_type: cDNA71  task: regulatory-profile72  text: ATGGAGCCGGCGGCGGGGAGCAGCATGGAGCCTTCGGCTGACTGGCTGGCCACGGCCGCGGCCCGGGGTCGGGTAGAGGAGGTGCGGGCGCTGCTGGAGGCGGGGGCGCTGCCCAACGCACCGAATAGTTACGGTCGGAGGCCGATCCAGGTCATGATGATGGGCAGCGCCCGAGTGGCGGAGCTGCTGCTGCTCCACGGCGCGGAGCCCAACTGCGCCGACCCCGCCACTCTCACCCGACCCGTGCACGACGCTGCCCGGGAGGGCTTCCTGGACACGCTGGTGGTGCTGCACCGGGCCGGGGCGCGGCTGGACGTGCGCGATGCCTGGGGCCGTCTGCCCGTGGACCTGGCTGAGGAGCTGGGCCATCGCGATGTCGCACGGTACCTGCGCGCGGCTGCGGGGGGCACCAGAGGCAGTAACCATGCCCGCATAGATGCCGCGGAAGGTCCCTCAGACATCCCCGATTGA73- example_title: human papillomavirus type 16 E674  pipeline_tag: regulatory-profile75  sequence_type: cDNA76  task: regulatory-profile77  text: ATGCACCAAAAGAGAACTGCAATGTTTCAGGACCCACAGGAGCGACCCAGAAAGTTACCACAGTTATGCACAGAGCTGCAAACAACTATACATGATATAATATTAGAATGTGTGTACTGCAAGCAACAGTTACTGCGACGTGAGGTATATGACTTTGCTTTTCGGGATTTATGCATAGTATATAGAGATGGGAATCCATATGCTGTATGTGATAAATGTTTAAAGTTTTATTCTAAAATTAGTGAGTATAGACATTATTGTTATAGTTTGTATGGAACAACATTAGAACAGCAATACAACAAACCGTTGTGTGATTTGTTAATTAGGTGTATTAACTGTCAAAAGCCACTGTGTCCTGAAGAAAAGCAAAGACATCTGGACAAAAAGCAAAGATTCCATAATATAAGGGGTCGGTGGACCGGTCGATGTATGTCTTGTTGCAGATCATCAAGAACACGTAGAGAAACCCAGCTGTAA78---79 80# BPNet81 82Base-resolution convolutional neural network for predicting transcription-factor binding profiles from DNA sequence.83 84## Disclaimer85 86This is an UNOFFICIAL implementation of [Base-resolution models of transcription-factor binding reveal soft motif syntax](https://doi.org/10.1038/s41588-021-00782-6) by Žiga Avsec, Melanie Weilert, et al.87 88The OFFICIAL repository of BPNet is at [kundajelab/bpnet](https://github.com/kundajelab/bpnet).89 90> [!TIP]91> The MultiMolecule team has confirmed that the provided model and checkpoints are producing the same intermediate representations as the original implementation.92 93**The team releasing BPNet did not write this model card for this model so this model card has been written by the MultiMolecule team.**94 95## Model Details96 97BPNet is a convolutional neural network (CNN) trained to predict base-resolution transcription-factor binding signal (ChIP-nexus) from primary DNA sequence. It uses a convolutional motif stem followed by a stack of dilated residual convolutions that aggregate ~1 kb of genomic context. The output is factorized into profile and count branches, and the usable base-resolution prediction is reconstructed by `BpNetForProfilePrediction.postprocess`. Please refer to the [Training Details](#training-details) section for more information on the training process.98 99### Model Specification100 101| Num Layers | Hidden Size | Num Parameters (M) | FLOPs (G) | MACs (G) |102| ---------- | ----------- | ------------------ | --------- | -------- |103| 10         | 64          | 0.13               | 0.24      | 0.12     |104 105### Links106 107- **Code**: [multimolecule.bpnet](https://github.com/DLS5-Omics/multimolecule/tree/master/multimolecule/models/bpnet)108- **Data**: [BPNet manuscript data](https://zenodo.org/records/4294904)109- **Paper**: [Base-resolution models of transcription-factor binding reveal soft motif syntax](https://doi.org/10.1038/s41588-021-00782-6)110- **Developed by**: Žiga Avsec, Melanie Weilert, Avanti Shrikumar, Sabrina Krueger, Amr Alexandari, Khyati Dalal, Robin Fropf, Charles McAnany, Julien Gagneur, Anshul Kundaje, Julia Zeitlinger111- **Model type**: 1D dilated CNN with factorized profile-and-count heads for base-resolution transcription-factor binding prediction112- **Original Repository**: [kundajelab/bpnet](https://github.com/kundajelab/bpnet)113 114## Usage115 116The model file depends on the [`multimolecule`](https://multimolecule.danling.org) library. You can install it using pip:117 118```bash119pip install multimolecule120```121 122### Direct Use123 124#### Transcription-Factor Binding Profile Prediction125 126You can use this model directly to predict transcription-factor binding profiles of a DNA sequence:127 128```python129>>> from multimolecule import DnaTokenizer, BpNetForProfilePrediction130 131>>> tokenizer = DnaTokenizer.from_pretrained("multimolecule/bpnet")132>>> model = BpNetForProfilePrediction.from_pretrained("multimolecule/bpnet")133>>> output = model(**tokenizer("ACGTNACGTN", return_tensors="pt"))134 135>>> output.keys()136odict_keys(['profile_logits', 'count_logits'])137 138>>> output["profile_logits"].shape139torch.Size([1, 10, 8])140 141>>> output["count_logits"].shape142torch.Size([1, 8])143 144>>> track = model.postprocess(output)145>>> track.shape146torch.Size([1, 10, 8])147```148 149The recombined `track` is the usable base-resolution prediction. The last dimension stacks `num_tasks` (Oct4, Sox2, Nanog, Klf4) by `num_strands` (forward, reverse).150 151### Interface152 153- **Input length**: 1000 bp DNA window154- **Output**: factorized `(profile_logits, count_logits)`; recombine the usable base-resolution track via `BpNetForProfilePrediction.postprocess`155- **Output shape**: `(batch_size, profile_length, num_tasks × num_strands)`; Oct4 / Sox2 / Nanog / Klf4 × forward / reverse = 8 channels156 157## Training Details158 159BPNet was trained to predict the base-resolution ChIP-nexus binding profiles of the pluripotency transcription factors Oct4, Sox2, Nanog and Klf4 in mouse embryonic stem cells.160 161### Training Data162 163The published BPNet-OSKN model was trained on ChIP-nexus profiles for Oct4, Sox2, Nanog and Klf4, using 1 kb genomic windows centered on detected binding peaks. The training regions and trained model files are distributed as part of the [BPNet manuscript data](https://zenodo.org/records/4294904).164 165### Training Procedure166 167#### Pre-training168 169The model was trained with a composite loss: a multinomial negative log-likelihood on the per-position profile shape plus a mean-squared-error regression on the log total counts.170 171- Optimizer: Adam172 173## Citation174 175```bibtex176@article{avsec2021baseresolution,177  author    = {Avsec, {\v{Z}}iga and Weilert, Melanie and Shrikumar, Avanti and Krueger, Sabrina and Alexandari, Amr and Dalal, Khyati and Fropf, Robin and McAnany, Charles and Gagneur, Julien and Kundaje, Anshul and Zeitlinger, Julia},178  title     = {Base-resolution models of transcription-factor binding reveal soft motif syntax},179  journal   = {Nature Genetics},180  volume    = 53,181  number    = 3,182  pages     = {354--366},183  year      = 2021,184  publisher = {Nature Publishing Group},185  doi       = {10.1038/s41588-021-00782-6}186}187```188 189> [!NOTE]190> The artifacts distributed in this repository are part of the MultiMolecule project.191> If MultiMolecule supports your research, please cite the MultiMolecule project as follows:192 193```bibtex194@software{chen_2024_12638419,195  author    = {Chen, Zhiyuan and Zhu, Sophia Y.},196  title     = {MultiMolecule},197  doi       = {10.5281/zenodo.12638419},198  publisher = {Zenodo},199  url       = {https://doi.org/10.5281/zenodo.12638419},200  year      = 2024,201  month     = may,202  day       = 4203}204```205 206## Contact207 208Please use GitHub issues of [MultiMolecule](https://github.com/DLS5-Omics/multimolecule/issues) for any questions or comments on the model card.209 210Please contact the authors of the [BPNet paper](https://doi.org/10.1038/s41588-021-00782-6) for questions or comments on the paper/model.211 212## License213 214This model implementation is licensed under the [GNU Affero General Public License](license.md).215 216For additional terms and clarifications, please refer to our [License FAQ](license-faq.md).217 218```spdx219SPDX-License-Identifier: AGPL-3.0-or-later220```