空间站广场

论文

Notebooks

比赛

课程

Apps

我的主页

我的Notebooks

我的论文库

我的足迹

我的工作空间

任务

节点

文件

数据集

镜像

项目

数据库

公开

基于Uni-Mol的分子&原子级别向量表征

Uni-Mol

Deep Learning

Uni-MolDeep Learning

Zhifeng Gao

Letian

更新于 2024-10-30

推荐镜像 :Basic Image:bohrium-notebook:2023-04-07

推荐机型 :c3_m4_1 * NVIDIA T4

基于Uni-Mol的分子向量表征

什么是分子/原子向量表征

导入Uni-Mol

基于Uni-Mol的分子向量表征

开源代码地址：https://github.com/deepmodeling/Uni-Mol
论文地址：https://openreview.net/forum?id=6K2RM6wVqKu

什么是分子/原子向量表征

分子/原子级别的向量表征是将原子或分子的化学和物理特性表示为数学向量的过程，在化学信息学、计算化学、药物设计和材料科学中具有重要意义。
原子级别的向量表征包括原子类型、电负性、原子半径、电子层结构、电荷分布和键类型等信息。例如，一个碳原子的向量可能是[6, 2.55, 0.77, [2, 4], 0.0, 1]。
分子级别的向量表征则涵盖分子结构信息、电子性质、几何结构、光谱性质、热力学性质和反应活性等。例如，一个水分子的向量可能是[18.015, 1.85, [0.957, 104.5], [0.34, 0.17, 1.85], -75.0]。

这种表征方法在机器学习模型、分子相似性搜索和化学反应预测等领域应用广泛，为定量分析和计算提供了重要基础，促进了化学和生物学的研究与应用。

代码

文本

导入Uni-Mol

代码

文本

Uni-Mol tools目前已经能够使用pip install来安装，运行下面的代码，一键安装。

代码

文本

[1]

!pip install unimol_tools

Looking in indexes: https://pypi.tuna.tsinghua.edu.cn/simple
Collecting unimol_tools
  Downloading https://pypi.tuna.tsinghua.edu.cn/packages/49/02/01b92f2a35425ccfd7675bf3ab6f0a45e6b0e9ff3e95c420ae062801af66/unimol_tools-0.1.0.post4-py3-none-any.whl (51 kB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 51.1/51.1 kB 1.3 MB/s eta 0:00:00
Collecting addict
  Downloading https://pypi.tuna.tsinghua.edu.cn/packages/6a/00/b08f23b7d7e1e14ce01419a467b583edbb93c6cdb8654e54a9cc579cd61f/addict-2.4.0-py3-none-any.whl (3.8 kB)
Requirement already satisfied: tqdm in /opt/conda/lib/python3.8/site-packages (from unimol_tools) (4.64.1)
Collecting rdkit
  Downloading https://pypi.tuna.tsinghua.edu.cn/packages/3d/84/63b2e66f5c7cb97ce994769afbbef85a1ac364fedbcb7d4a3c0f15d318a5/rdkit-2024.3.5-cp38-cp38-manylinux_2_28_x86_64.whl (33.1 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 33.1/33.1 MB 2.7 MB/s eta 0:00:0000:0100:01
Requirement already satisfied: torch in /opt/conda/lib/python3.8/site-packages (from unimol_tools) (1.13.1+cu116)
Requirement already satisfied: numpy<2.0.0,>=1.22.4 in /opt/conda/lib/python3.8/site-packages (from unimol_tools) (1.22.4)
Requirement already satisfied: pandas<2.0.0 in /opt/conda/lib/python3.8/site-packages (from unimol_tools) (1.5.3)
Requirement already satisfied: scikit-learn in /opt/conda/lib/python3.8/site-packages (from unimol_tools) (1.0.2)
Requirement already satisfied: joblib in /opt/conda/lib/python3.8/site-packages (from unimol_tools) (1.2.0)
Requirement already satisfied: pyyaml in /opt/conda/lib/python3.8/site-packages (from unimol_tools) (6.0)
Requirement already satisfied: python-dateutil>=2.8.1 in /opt/conda/lib/python3.8/site-packages (from pandas<2.0.0->unimol_tools) (2.8.2)
Requirement already satisfied: pytz>=2020.1 in /opt/conda/lib/python3.8/site-packages (from pandas<2.0.0->unimol_tools) (2022.7)
Requirement already satisfied: Pillow in /opt/conda/lib/python3.8/site-packages (from rdkit->unimol_tools) (9.4.0)
Requirement already satisfied: scipy>=1.1.0 in /opt/conda/lib/python3.8/site-packages (from scikit-learn->unimol_tools) (1.7.3)
Requirement already satisfied: threadpoolctl>=2.0.0 in /opt/conda/lib/python3.8/site-packages (from scikit-learn->unimol_tools) (3.1.0)
Requirement already satisfied: typing-extensions in /opt/conda/lib/python3.8/site-packages (from torch->unimol_tools) (4.5.0)
Requirement already satisfied: six>=1.5 in /opt/conda/lib/python3.8/site-packages (from python-dateutil>=2.8.1->pandas<2.0.0->unimol_tools) (1.16.0)
Installing collected packages: addict, rdkit, unimol_tools
Successfully installed addict-2.4.0 rdkit-2024.3.5 unimol_tools-0.1.0.post4
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv

代码

文本

[3]

#考虑库不兼容的问题，这里先升级一下numpy

!pip install --upgrade numpy

Looking in indexes: https://pypi.tuna.tsinghua.edu.cn/simple
Requirement already satisfied: numpy in /opt/conda/lib/python3.8/site-packages (1.24.4)
WARNING: Running pip as the 'root' user can result in broken permissions and conflicting behaviour with the system package manager. It is recommended to use a virtual environment instead: https://pip.pypa.io/warnings/venv

代码

文本

因为需要从hugging face上下载，这里先配置一下ip

代码

文本

[4]

import os

os.environ['HTTP_PROXY'] = 'http://ga.dp.tech:8118'

os.environ['HTTPS_PROXY'] = 'http://ga.dp.tech:8118'

代码

文本

[5]

from unimol_tools import UniMolRepr

import numpy as np

import pandas as pd

# single smiles unimol representation

clf = UniMolRepr(data_type='molecule', remove_hs=False)

smiles = 'c1ccc(cc1)C2=NCC(=O)Nc3c2cc(cc3)[N+](=O)[O]'

smiles_list = [smiles]

unimol_repr = clf.get_repr(smiles_list, return_atomic_reprs=True)

# CLS token repr

print(np.array(unimol_repr['cls_repr']).shape)

# atomic level repr, align with rdkit mol.GetAtoms()

print(np.array(unimol_repr['atomic_reprs']).shape)

/opt/conda/lib/python3.8/site-packages/scipy/__init__.py:146: UserWarning: A NumPy version >=1.16.5 and <1.23.0 is required for this version of SciPy (detected version 1.24.4
  warnings.warn(f"A NumPy version >={np_minversion} and <{np_maxversion}"
2024-10-30 16:58:52 | unimol_tools/weights/weighthub.py | 17 | INFO | Uni-Mol Tools | Weights will be downloaded to default directory: /opt/conda/lib/python3.8/site-packages/unimol_tools/weights
2024-10-30 16:58:53 | unimol_tools/weights/weighthub.py | 33 | INFO | Uni-Mol Tools | Downloading mol_pre_all_h_220816.pt

2024-10-30 16:59:15 | unimol_tools/weights/weighthub.py | 33 | INFO | Uni-Mol Tools | Downloading mol.dict.txt

2024-10-30 16:59:17 | unimol_tools/models/unimol.py | 120 | INFO | Uni-Mol Tools | Loading pretrained weights from /opt/conda/lib/python3.8/site-packages/unimol_tools/weights/mol_pre_all_h_220816.pt
2024-10-30 16:59:21 | unimol_tools/data/conformer.py | 89 | INFO | Uni-Mol Tools | Start generating conformers...
1it [00:00, 13.71it/s]
2024-10-30 16:59:21 | unimol_tools/data/conformer.py | 93 | INFO | Uni-Mol Tools | Succeed to generate conformers for 100.00% of molecules.
2024-10-30 16:59:21 | unimol_tools/data/conformer.py | 95 | INFO | Uni-Mol Tools | Succeed to generate 3d conformers for 100.00% of molecules.
100%|██████████| 1/1 [00:01<00:00,  1.18s/it](1, 512)
(1, 32, 512)

代码

文本

[6]

%%bash

# 下载样例数据, CNS drug data

rm -rf mol_train.csv

wget -nv https://bohrium-example.oss-cn-zhangjiakou.aliyuncs.com/unimol-qsar/mol_train.csv

2024-10-30 16:59:26 URL:https://bohrium-example.oss-cn-zhangjiakou.aliyuncs.com/unimol-qsar/mol_train.csv [30600/30600] -> "mol_train.csv" [1]

代码

文本

[7]

smiles_list = pd.read_csv('mol_train.csv')['SMILES'].to_list()

y = pd.read_csv('mol_train.csv')['TARGET'].to_list()

repr_dict = clf.get_repr(smiles_list)

unimol_repr_list = np.array(repr_dict['cls_repr'])

2024-10-30 16:59:31 | unimol_tools/data/conformer.py | 89 | INFO | Uni-Mol Tools | Start generating conformers...
700it [00:39, 17.85it/s] 
2024-10-30 17:00:10 | unimol_tools/data/conformer.py | 93 | INFO | Uni-Mol Tools | Succeed to generate conformers for 100.00% of molecules.
2024-10-30 17:00:10 | unimol_tools/data/conformer.py | 95 | INFO | Uni-Mol Tools | Succeed to generate 3d conformers for 100.00% of molecules.
100%|██████████| 22/22 [00:07<00:00,  2.96it/s]

代码

文本

[8]

print(unimol_repr_list.shape)

(700, 512)

代码

文本

[9]

import numpy as np

import matplotlib.pyplot as plt

from mpl_toolkits.mplot3d import Axes3D

from sklearn.decomposition import PCA

代码

文本

[10]

pca = PCA(n_components=2)

X_reduced = pca.fit_transform(unimol_repr_list)

代码

文本

[11]

# 可视化

colors = ['r', 'g', 'b']

markers = ['s', 'o', 'D']

labels = ['Target:0','Target:1']

plt.figure(figsize=(8, 6))

for label, color, marker in zip(np.unique(y), colors, markers):

plt.scatter(X_reduced[y == label, 0],

X_reduced[y == label, 1],

c=color,

marker=marker,

label=labels[label],

edgecolors='black')

plt.xlabel('Principal Component 1')

plt.ylabel('Principal Component 2')

plt.legend(loc='best')

plt.title('Unimol Repr')

plt.show()

代码

文本

[ ]

代码

文本

Uni-Mol

Deep Learning

Uni-MolDeep Learning

已赞4

本文被以下合集收录

Uni-Mol最佳实践

Zhifeng Gao

更新于 2024-10-28

5 篇7 人关注

uni-mol

Nshima

更新于 2024-01-23

1 篇0 人关注