introvoyz041/vesm_scores

---
language:
- en
tags:
- biology
- ESM
- language-model
- protein
- VEP
pretty_name: VESM scores
size_categories:
- 100M<n<1B
---

# Proteome-wide VESM variant effect scores

This repository provides precomputed **proteome-wide (UniProtKB, hg19, and hg38) variant-effect prediction scores** using the latest VESM models developed in the paper ["Compressing the collective knowledge of ESM into a single protein language model"](vesm_arxiv) by Tuan Dinh, Seon-Kyeong Jang, Noah Zaitlen and Vasilis Ntranos.
- **Models:** VESM_3B, VESM3, sequence-only VESM3, and VESM++ (available at https://huggingface.co/ntranoslab/vesm).

```VESM_3B and VESM3 are individual protein language models based on ESM2 (3B) and ESM3. The sequence-only VESM3 is the version of VESM3 using only sequence as the input. VESM++ is the ensemble of VESM_3B and VESM3. ```

Please see the corresponding GitHub repo (https://github.com/ntranoslab/vesm) for more details.

## License <a name="license"></a>
The predictions of VESM_3B are distributed under the MIT License.
The VESM3 and VESM++ models are built with ESM3-Open (EvolutionaryScale), which is available under a [non-commercial license agreement](https://www.evolutionaryscale.ai/policies/cambrian-open-license-agreement).

---
语言：英语
标签：生物学、ESM、语言模型、蛋白质、VEP
展示名称：VESM分数
规模类别：1亿 < 样本量 < 10亿
---

# 全蛋白质组VESM变异效应分数数据集

本数据集仓库提供了由Tuan Dinh、Seon-Kyeong Jang、Noah Zaitlen与Vasilis Ntranos在论文《将ESM的集体知识压缩至单一蛋白质语言模型（protein language model）》（链接：vesm_arxiv）中开发的最新VESM模型，所预计算得到的**全蛋白质组（通用蛋白质知识库(UniProtKB)、hg19与hg38）变异效应预测分数**。
- **模型列表**：VESM_3B、VESM3、仅以序列为输入的VESM3以及VESM++（模型获取链接：https://huggingface.co/ntranoslab/vesm）。

VESM_3B与VESM3均为基于ESM2（3B）与ESM3的独立蛋白质语言模型。仅以序列为输入的VESM3为仅使用序列作为输入的VESM3变体。VESM++则为VESM_3B与VESM3的集成模型。

更多详细信息请参阅对应GitHub仓库（https://github.com/ntranoslab/vesm）。

## 许可证 <a name="license"></a>
VESM_3B的预测结果以MIT许可证进行分发。
VESM3与VESM++模型基于ESM3-Open（EvolutionaryScale）构建，该模型采用[非商业性许可协议](https://www.evolutionaryscale.ai/policies/cambrian-open-license-agreement)进行授权。