{
  "id": 663136,
  "title": "Has anyone tried this method: DeepGO-SE",
  "url": "/competitions/cafa-6-protein-function-prediction/discussion/663136",
  "author_name": "",
  "post_date": "2025-12-16T15:35:56.859015200Z",
  "votes": 4,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Has anyone tried this method published in Nature last year: <a href=\"https://www.nature.com/articles/s42256-024-00795-w\" target=\"_blank\">https://www.nature.com/articles/s42256-024-00795-w</a></p>\n<p>Essentially, they do the following:</p>\n<ul>\n<li>start from ESM2 embeddings for protein sequences</li>\n<li>extend these embeddings using an MLP</li>\n<li>they encode GO terms using a network approach (if I understood it right)</li>\n</ul>\n<p><img src=\"https://media.springernature.com/full/springer-static/image/art%3A10.1038%2Fs42256-024-00795-w/MediaObjects/42256_2024_795_Fig1_HTML.png\" alt=\"DeepGO - SE fig 1\"></p>",
  "messages": [
    {
      "id": "3377586",
      "postDate": "12/16/2025 15:35:56",
      "content": "<p>Has anyone tried this method published in Nature last year: <a href=\"https://www.nature.com/articles/s42256-024-00795-w\" target=\"_blank\">https://www.nature.com/articles/s42256-024-00795-w</a></p>\n<p>Essentially, they do the following:</p>\n<ul>\n<li>start from ESM2 embeddings for protein sequences</li>\n<li>extend these embeddings using an MLP</li>\n<li>they encode GO terms using a network approach (if I understood it right)</li>\n</ul>\n<p><img src=\"https://media.springernature.com/full/springer-static/image/art%3A10.1038%2Fs42256-024-00795-w/MediaObjects/42256_2024_795_Fig1_HTML.png\" alt=\"DeepGO - SE fig 1\"></p>",
      "rawMarkdown": "Has anyone tried this method published in Nature last year: https://www.nature.com/articles/s42256-024-00795-w\n\nEssentially, they do the following:\n- start from ESM2 embeddings for protein sequences\n- extend these embeddings using an MLP\n- they encode GO terms using a network approach (if I understood it right)\n\n![DeepGO - SE fig 1](https://media.springernature.com/full/springer-static/image/art%3A10.1038%2Fs42256-024-00795-w/MediaObjects/42256_2024_795_Fig1_HTML.png)",
      "votes": null
    },
    {
      "id": "3377999",
      "postDate": "12/17/2025 10:14:18",
      "content": "<p>Likely, given they are on the LB :-D</p>",
      "rawMarkdown": "Likely, given they are on the LB :-D",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3377999,
      "author_name": "mtinti",
      "author_url": "",
      "post_date": "12/17/2025 10:14:18",
      "content": "<p>Likely, given they are on the LB :-D</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3377586": "Has anyone tried this method published in Nature last year: https://www.nature.com/articles/s42256-024-00795-w\n\nEssentially, they do the following:\n- start from ESM2 embeddings for protein sequences\n- extend these embeddings using an MLP\n- they encode GO terms using a network approach (if I understood it right)\n\n![DeepGO - SE fig 1](https://media.springernature.com/full/springer-static/image/art%3A10.1038%2Fs42256-024-00795-w/MediaObjects/42256_2024_795_Fig1_HTML.png)",
    "3377999": "Likely, given they are on the LB :-D"
  },
  "source": "meta"
}