Bert Protein Classifier

🚀 蛋白质多标签分类模型

本项目是一个基于微调的Bert-Base-Uncased模型，主要用于多标签分类任务。该模型能够根据蛋白质的氨基酸序列来预测其功能，为生物学研究和蛋白质分析提供了有力支持。

🚀 快速开始

本模型的使用方式很简单，在推理框中粘贴蛋白质序列，模型就会输出该序列与某些GO术语相关联的概率。

例如，输入以下蛋白质序列： MMSTTHLLVFLLGVVTLTTPTFGTYESPNYGKPPTPVFKPPKVKPPPYEPKPPVYEPPKKEKPEPKPPVYAPPKKEKHGPKPTMYEPPKKEKPEPKPPVYTPPKKEVPKPKPPVYEPPKKEKPEPKPPIYTPPKKEKPEPKPPVYEPPKKEKPEPKPPVYTPPKKEKPEPKPPVYEPPKKPPMYEPKPPKPPVYTPPKKEKPEPKPPMYEPPKKPPMYEPKPPKPPVYTPPKKEKPEPKPPMYQPPNNPPIYEPKPPKPPVYAPPKEEKPKPKPPVYEPPAHEPPYGHYPGHPPLGKPQ

模型将输出如下分数：

[
  [
    {
      "label": "GO:0000122",
      "score": 0.29775485396385193
    },
    {
      "label": "GO:0000070",
      "score": 0.10477513074874878
    },
    {
      "label": "GO:0000075",
      "score": 0.08593793958425522
    },
    {
      "label": "GO:0000118",
      "score": 0.05860009789466858
    },
    {
      "label": "GO:0000082",
      "score": 0.05373986065387726
    },
    {
      "label": "GO:0000077",
      "score": 0.03928716108202934
    },
    {
      "label": "GO:0000096",
      "score": 0.03705739229917526
    },
    {
      "label": "GO:0000079",
      "score": 0.02797592058777809
    },
    {
      "label": "GO:0000045",
      "score": 0.026528609916567802
    },
    {
      "label": "GO:0000097",
      "score": 0.026119187474250793
    },
    {
      "label": "GO:0000086",
      "score": 0.019697198644280434
    },
    {
      "label": "GO:0000049",
      "score": 0.018551582470536232
    },
    {
      "label": "GO:0000041",
      "score": 0.016929756850004196
    },
    {
      "label": "GO:0000054",
      "score": 0.015105823054909706
    },
    {
      "label": "GO:0000083",
      "score": 0.01434631273150444
    },
    {
      "label": "GO:0000105",
      "score": 0.013960960321128368
    },
    {
      "label": "GO:0000076",
      "score": 0.013064960949122906
    },
    {
      "label": "GO:0000109",
      "score": 0.012523632496595383
    },
    {
      "label": "GO:0000113",
      "score": 0.012152223847806454
    },
    {
      "label": "GO:0000062",
      "score": 0.01127714291214943
    },
    {
      "label": "GO:0000101",
      "score": 0.011041304096579552
    }
]

📚 详细文档

模型信息

这是一个经过微调的Bert-Base-Uncased模型，用于多标签分类任务。该任务主要是基于蛋白质的氨基酸序列来预测其功能。模型以序列数据和蛋白质类别名称作为输入，并输出概率分数，即该序列属于某个类别的可能性。