Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Co-evolution Transformer for Protein Contact Prediction
Authors: He Zhang, Fusong Ju, Jianwei Zhu, Liang He, Bin Shao, Nanning Zheng, Tie-Yan Liu
NeurIPS 2021 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on two rigorous benchmark datasets demonstrate the effectiveness of Co T. |
| Researcher Affiliation | Collaboration | He Zhang Xi an Jiaotong University EMAIL Fusong Ju Microsoft Research Asia EMAIL |
| Pseudocode | No | No pseudocode or clearly labeled algorithm blocks were found in the paper. |
| Open Source Code | Yes | Our code will be released at https://github.com/microsoft/Protein Folding/tree/main/ coevolution_transformer. |
| Open Datasets | Yes | All models are trained on 96, 167 protein structures (chains) collected from PDB [37] (before Apr. 1, 2020) |
| Dataset Splits | Yes | All models are trained on 96, 167 protein structures (chains) collected from PDB [37] (before Apr. 1, 2020), which are split into the train and validation sets (95, 667 and 500 proteins, respectively). |
| Hardware Specification | Yes | The total training cost of the Co T model is about 30 hours on 4 Tesla V100 GPU cards. |
| Software Dependencies | No | The paper mentions 'Adam optimizer [38]' but does not provide specific software names with version numbers for reproducibility (e.g., Python, PyTorch, TensorFlow versions). |
| Experiment Setup | Yes | The Co T model is equipped with 6 Co T layers with hidden size as 128 and the attention head number as 8. All models are trained with Adam optimizer [38] via a cross-entropy loss for 100k iterations. The learning rate, the weight decay, and the batch size are set to 10 4, 0.01, and 16 respectively. |