{
  "id": 570003,
  "title": "Paper: MSAGPT: Neural Prompting Protein Structure Prediction via MSA Generative Pre-Training",
  "url": "/competitions/stanford-rna-3d-folding/discussion/570003",
  "author_name": "",
  "post_date": "2025-03-25T11:52:05.344470500Z",
  "votes": 6,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Paper: MSAGPT: Neural Prompting Protein Structure Prediction via MSA Generative Pre-Training</p>\n<p>MSAGPT is a powerful protein language model (PLM). MSAGPT has 3 billion parameters with three versions of the model, MSAGPT, MSAGPT-Sft, and MSAGPT-Dpo, supporting zero-shot and few-shot MSA generation.</p>\n<p>MSAGPT achieves state-of-the-art structural prediction performance on natural MSA-scarce scenarios.<br>\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F99fcdeee1ace7d91ef1422595bf02e66%2FSelection_146.png?generation=1742903519580061&amp;alt=media\" alt=\"\"></p>\n<p>can this be used?<br>\n<a href=\"https://huggingface.co/genbio-ai/AIDO.RNA-1.6B\" target=\"_blank\">https://huggingface.co/genbio-ai/AIDO.RNA-1.6B</a><br>\n<a href=\"https://huggingface.co/genbio-ai/AIDO.RNA-650M\" target=\"_blank\">https://huggingface.co/genbio-ai/AIDO.RNA-650M</a></p>",
  "messages": [
    {
      "id": "3159237",
      "postDate": "03/25/2025 11:52:05",
      "content": "<p>Paper: MSAGPT: Neural Prompting Protein Structure Prediction via MSA Generative Pre-Training</p>\n<p>MSAGPT is a powerful protein language model (PLM). MSAGPT has 3 billion parameters with three versions of the model, MSAGPT, MSAGPT-Sft, and MSAGPT-Dpo, supporting zero-shot and few-shot MSA generation.</p>\n<p>MSAGPT achieves state-of-the-art structural prediction performance on natural MSA-scarce scenarios.<br>\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F99fcdeee1ace7d91ef1422595bf02e66%2FSelection_146.png?generation=1742903519580061&amp;alt=media\" alt=\"\"></p>\n<p>can this be used?<br>\n<a href=\"https://huggingface.co/genbio-ai/AIDO.RNA-1.6B\" target=\"_blank\">https://huggingface.co/genbio-ai/AIDO.RNA-1.6B</a><br>\n<a href=\"https://huggingface.co/genbio-ai/AIDO.RNA-650M\" target=\"_blank\">https://huggingface.co/genbio-ai/AIDO.RNA-650M</a></p>",
      "rawMarkdown": "Paper: MSAGPT: Neural Prompting Protein Structure Prediction via MSA Generative Pre-Training\n\nMSAGPT is a powerful protein language model (PLM). MSAGPT has 3 billion parameters with three versions of the model, MSAGPT, MSAGPT-Sft, and MSAGPT-Dpo, supporting zero-shot and few-shot MSA generation.\n\nMSAGPT achieves state-of-the-art structural prediction performance on natural MSA-scarce scenarios.\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F99fcdeee1ace7d91ef1422595bf02e66%2FSelection_146.png?generation=1742903519580061&alt=media)\n\n\ncan this be used?\nhttps://huggingface.co/genbio-ai/AIDO.RNA-1.6B\nhttps://huggingface.co/genbio-ai/AIDO.RNA-650M",
      "votes": null
    },
    {
      "id": "3159439",
      "postDate": "03/25/2025 15:39:41",
      "content": "<p>AIDO.RNA was pre-trained using a masked language modeling objective. A simple strategy is to randomly mask some of the tokens for the query sequence and predict the masked tokens to get evolutionary similar sequences. However, the MSA generated in this way could only consider mutations, no insertion or deletions.</p>",
      "rawMarkdown": "AIDO.RNA was pre-trained using a masked language modeling objective. A simple strategy is to randomly mask some of the tokens for the query sequence and predict the masked tokens to get evolutionary similar sequences. However, the MSA generated in this way could only consider mutations, no insertion or deletions.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3159439,
      "author_name": "zoushuxian",
      "author_url": "",
      "post_date": "03/25/2025 15:39:41",
      "content": "<p>AIDO.RNA was pre-trained using a masked language modeling objective. A simple strategy is to randomly mask some of the tokens for the query sequence and predict the masked tokens to get evolutionary similar sequences. However, the MSA generated in this way could only consider mutations, no insertion or deletions.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3159237": "Paper: MSAGPT: Neural Prompting Protein Structure Prediction via MSA Generative Pre-Training\n\nMSAGPT is a powerful protein language model (PLM). MSAGPT has 3 billion parameters with three versions of the model, MSAGPT, MSAGPT-Sft, and MSAGPT-Dpo, supporting zero-shot and few-shot MSA generation.\n\nMSAGPT achieves state-of-the-art structural prediction performance on natural MSA-scarce scenarios.\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F113660%2F99fcdeee1ace7d91ef1422595bf02e66%2FSelection_146.png?generation=1742903519580061&alt=media)\n\n\ncan this be used?\nhttps://huggingface.co/genbio-ai/AIDO.RNA-1.6B\nhttps://huggingface.co/genbio-ai/AIDO.RNA-650M",
    "3159439": "AIDO.RNA was pre-trained using a masked language modeling objective. A simple strategy is to randomly mask some of the tokens for the query sequence and predict the masked tokens to get evolutionary similar sequences. However, the MSA generated in this way could only consider mutations, no insertion or deletions."
  },
  "source": "meta"
}