Help
Guides and instructions
User Guide
CoPo (Covalent Pocket) is a novel multimodal deep learning framework for accurately identifying covalent pockets in protein structures. Inputting protein structural data, users will obtain precise predictions of binding sites and interpretable residue-level insights into pocket binding and reactivity by integrating sequence, structure, and surface features. We have evaluated CoPo on an independent dataset of over one hundred covalent complexes, and achieved a Matthews correlation coefficient (MCC) of 0.7298, significantly outperforming conventional algorithms and existing state-of-the-art methods. CoPo will facilitate the covalent drug discovery for diverse targets, including kinases, traditionally "undruggable" targets, and non-cysteine proteins.
The main functional module of CoPo platform is the Detect webpage. In Home webpage, users can click the Get Started button or the Detect item in the navigation bar to jump to the Detect webpage.
(1) A protein structure file in PDB format is required as input for computation purpose. Users can upload one PDB file, with a size no more than 10MB, every time in the Detect webpage.
(2) Users also can name their jobs by their preferences and habits (This step is optional).
(3) Users also can enter their email address (This step is optional).
(4) When users click the Submit button, job information will be submitted to the back-end server. CoPo platform will start computing as soon as possible.
(5) If users need to check the example, users can click the View Example button to view a pre-calculated example case.
After users submit their job, CoPo platform will calculate and show the prediction results as soon as possible. The results can be divided into the following 4 parts, and all results can be downloaded as PDB and CSV files.
(1) Basic Information: Input PDB file name, job status, and calculation duration.
(2) Interactive 3D Visualization: 3D protein structure with predicted reaction sites (Red) and binding sites (Green) highlighted.
(3) Detailed Prediction Table: A searchable list of predicted residues with their type and confidence scores.
(4) Downloads: The processed PDB structure and full prediction results in CSV format.
- Data Privacy: We respect your research data. All uploaded files and prediction results are stored securely and are only accessible via the unique Job ID (and password, if set).
- Data Retention: To maintain server performance, all job data will be automatically deleted from the server 7 days after creation. Please ensure you download your results within this period.
- Browser Compatibility: The 3D visualization module relies on WebGL. For the best experience, we recommend using modern browsers such as Chrome, Firefox, or Edge.
- Citation: If you find CoPo useful for your research, please cite our work as listed on the Publication page.