摘要:本文以化合物IP4为例,演示了如何用DP4-AI的工作流自动进行立体化学结构确证。

作者:肖高铿
日期:2020-10-18

关于DP4-AI

在前文DP4-AI自动NMR数据分析:直接从光谱到结构中,我们详细介绍了DP4-AI。虽然有AI字样,但与目前火热的AI概念并无任何关系。相反它是基于物理的方法进行NMR预测,然后通过比较理论NMR与实验NMR来识别可能的差向异构体。假设你有NMR原始数据(13C-NMR和/或1H-NMR数据),你可以通过DP4概率的计算来识别您的化合物可能是差向异构体中的哪一个。本文将通过一个算例演示如何使用DP4-AI进行全自动的DP4计算。

DP4-AI下载

下载:https://github.com/KristapsE/DP4-AI

或者通过git:

1
git clone https://github.com/KristapsE/DP4-AI.git

算例介绍

IP4化合物结构

Figure 1. 算例化合物IP4

本算例化合物IP4的结构如图1所示,已经准备好了3D的SDF文件、13C-NMR和/或1H-NMR原始数据(下载),这些文件保存在同一目录下,结构如下:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
demo
├── IP4_.sdf
└── NMR_folder
    ├── Carbon
    │   ├── acqu
    │   ├── acqus
    │   ├── audita.txt
    │   ├── cpdprg2
    │   ├── fid
    │   ├── format.temp
    │   ├── fq1list
    │   ├── pdata
    │   │   └── 1
    │   │       ├── 1i
    │   │       ├── 1r
    │   │       ├── auditp.txt
    │   │       ├── outd
    │   │       ├── proc
    │   │       ├── procs
    │   │       ├── thumb.png
    │   │       └── title
    │   ├── pulseprogram
    │   ├── scon2
    │   ├── shimvalues
    │   ├── specpar
    │   ├── uxnmr.info
    │   └── uxnmr.par
    └── Proton
        ├── acqu
        ├── acqus
        ├── audita.txt
        ├── fid
        ├── format.temp
        ├── fq1list
        ├── pdata
        │   └── 1
        │       ├── 1i
        │       ├── 1r
        │       ├── auditp.txt
        │       ├── outd
        │       ├── proc
        │       ├── procs
        │       ├── thumb.png
        │       └── title
        ├── pulseprogram
        ├── scon2
        ├── shimvalues
        ├── specpar
        ├── uxnmr.info
        └── uxnmr.par
 
7 directories, 42 files

你需要下载该文件到您的一个目录下,解压,并切换到该目录。到此,你已经准备好开始DP4-AI计算了。

DP4-AI的使用

1.使用方法

键入PyDP4 -h命令获得帮助:

1
 python /<安装路径>/DP4-AI/PyDP4.py -h

其中“安装路径”用你实际的路径代替,键入命令后出现帮助提示,按提示完成指令键入。

2. 起始的文件

以作者用到化合物IP4为例,以说明DP4计算的流程。初始的输入信息包括:

  1. IP4的3D结构
  2. IP4的3D结构以MDL SDF格式保存为IP4_.sdf

  3. NMR实验数据
  4. 原始的13与C1H-NMR实验数据分别以Carbon与Proton命名,保存在目录NMR_Folder里。

3. 使用gmns工作流

现在,可以开始用DP4-AI的自动工作流开始进行DP4计算。我们先试试作者的MM方法(直接用分子力场的构象进行NMR计算)。该策略使用工作流(-w gmns)来实现,包含了如下计算过程:差向异构体的枚举(g)、MMFF力场构象搜索(m)、NMR核磁计算、计算的NMR与实验NMR的数据提取与DP4计算。命令如下:

1
PyDP4.py -s chloroform -w gmns -m t -f mmff IP4_.sdf NMR_folder

其中-s chloroform是告诉DP4-AI在进行NMR计算时需要考虑溶剂(氯仿)的效应;-m t的意思是用Tinker进行构象搜索。

结果如下:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
==========================
PyDP4 script,
integrating Tinker/MacroModel,
Gaussian/NWChem and DP4
v1.0
 
Copyright (c) 2015-2019 Kristaps Ermanis, Alexander Howarth, Jonathan M. Goodman
Distributed under MIT license
==========================
 
Settings read from settings.cfg:
  TinkerPath: /home/cloudam/software/tinker
  NWChemPath: /public/software/.local/easybuild/software/NWChem/7.0.0-intel-2019b-Python-3.7.4/bin/nwchem
  GausPath: g16
['IP4_.sdf']
NMR_folder
NMR_path
/home/cloudam/work/dp4/IP4_tinker/NMR_folder
Current working directory: /home/cloudam/work/dp4/IP4_tinker
Initial input files: ['IP4_.sdf']
NMR file: [PosixPath('/home/cloudam/work/dp4/IP4_tinker/NMR_folder/Carbon'), PosixPath('/home/cloudam/work/dp4/IP4_tinker/NMR_folder/Proton')]
Workflow: gmns
 
Generating diastereomers...
Getting inchi from file  IP4_.sdf
Getting inchi from file  IP4_.sdf
Number of diastereomers to be generated: 2
Isomer 0 inchi = InChI=1S/C6H10O/c1-5-3-2-4-6(5)7/h2-3,5-7H,4H2,1H3/t5-,6-/m0/s1/f/
Isomer 1 inchi = InChI=1S/C6H10O/c1-5-3-2-4-6(5)7/h2-3,5-7H,4H2,1H3/t5-,6+/m0/s1/f/
Generated input files: ['IP4_1', 'IP4_2']
 
Assuming all computations are done? ...  False
Using preexisting DFT data? ...  False
Performing conformational search using Tinker
 
Setting up Tinker files...
Tinker input for IP4_1 prepared.
Tinker input for IP4_2 prepared.
 
Running Tinker...
Output files for IP4_1 already exist, skipping.
Output files for IP4_2 already exist, skipping.
IP4_1 is matching conformational search output for IP4_1
Reading IP4_1
Number of accepted conformers by energies: 3
IP4_2 IP4_1 Nope
IP4_1 IP4_2 Nope
IP4_2 is matching conformational search output for IP4_2
Reading IP4_2
Number of accepted conformers by energies: 5
 
Pruning conformers...
IP4_1: 3 conformers after pruning with 0A RMSD cutoff
IP4_2: 5 conformers after pruning with 0A RMSD cutoff
 
Setting up NMR calculations...
 
Running NMR calculations...
 
Running Gaussian DFT NMR calculations locally...
There were no jobs to run.
 
Reading data from the output files...
IP4_1ginp001.out 17
IP4_1ginp002.out 17
IP4_1ginp003.out 17
IP4_2ginp001.out 17
IP4_2ginp002.out 17
IP4_2ginp003.out 17
IP4_2ginp004.out 17
IP4_2ginp005.out 17
Shieldings:
IP4_1:
[52.7725, 142.6171, 106.6767, 137.7624, 44.3868, 262.8271, 169.0847, 29.3631, 29.358, 30.1239, 27.8744, 26.3857, 26.1914, 31.7494, 31.5088, 30.8564, 30.9325]
[51.1922, 147.0432, 106.3846, 133.0215, 45.7752, 264.1163, 169.3307, 29.4695, 29.7625, 29.6988, 27.8628, 26.3143, 26.2597, 31.6062, 31.5343, 30.9341, 31.0055]
[52.1369, 144.3506, 107.3824, 134.4598, 45.1125, 256.7895, 169.721, 29.3409, 29.4651, 29.8111, 28.2077, 26.3773, 26.2689, 30.9988, 31.5238, 30.8044, 30.9704]
IP4_2:
[52.5388, 147.5788, 111.0206, 142.2785, 45.1509, 290.3178, 172.6921, 29.6835, 29.5192, 29.4242, 27.249, 26.4213, 26.3882, 31.3379, 31.1798, 31.191, 31.1351]
[53.7533, 143.4332, 112.0511, 142.9355, 44.3782, 280.9794, 172.8862, 29.8359, 29.2278, 29.4755, 27.2404, 26.4982, 26.3565, 31.3338, 31.2015, 31.1403, 30.7387]
[51.0884, 141.8932, 111.8456, 140.8611, 45.4508, 282.0051, 173.2361, 29.7459, 29.3937, 29.3827, 28.0109, 26.3166, 26.4308, 31.3822, 30.8604, 31.2061, 30.909]
[50.699, 144.5118, 112.3287, 140.6304, 45.8268, 292.835, 175.1699, 29.5901, 29.4615, 29.4456, 27.7753, 26.259, 26.4269, 31.5163, 30.8012, 31.0402, 31.2906]
[52.0486, 140.3173, 112.6003, 140.7325, 44.5733, 275.5862, 174.1799, 30.0199, 29.3096, 29.4031, 27.7075, 26.3135, 26.3433, 31.8847, 30.9336, 30.8427, 30.8827]
Energies:
IP4_1: [-309.84674683, -309.847000007, -309.847961686]
IP4_2: [-309.846569715, -309.845282065, -309.847936164, -309.847324941, -309.845372646]
 
Setting TMS computational NMR shielding constant references
Setting TMS references to 188.452125 and 32.1243166667
 
Converting DFT data to NMR shifts...
[0.16868446982677085, 0.22056053309520496, 0.6107549970780243]
[0.12478688500156183, 0.03190761049375939, 0.5305037184618006, 0.277681469399478, 0.03512031664340019]
WARNING: NMR shift calculation currently ignores the instruction to exclude atoms from analysis
C shifts for isomer 0:
136.442, 43.808, 81.424, 53.763, 143.343, 18.928
H shifts for isomer 0:
2.751, 2.612, 2.285, 4.049, 5.760, 5.871, 0.865, 0.601, 1.283, 1.153
C shifts for isomer 1:
137.198, 45.137, 76.557, 47.425, 143.026, 14.728
H shifts for isomer 1:
2.417, 2.704, 2.715, 4.309, 5.805, 5.706, 0.694, 1.227, 0.981, 1.088
 
Reading experimental NMR data...
[PosixPath('/home/cloudam/work/dp4/IP4_tinker/NMR_folder/Carbon'), PosixPath('/home/cloudam/work/dp4/IP4_tinker/NMR_folder/Proton')]
Cshifts [16877, 24398, 27885, 35275, 36948, 41526, 44704, 45401, 48935, 49415, 49494, 50315, 51088, 51312, 52165, 52748, 52866, 52924, 52967, 53041, 53446, 53492, 53797, 53892, 54317, 54566, 56173, 56629, 56817, 56866, 56916, 57006, 57260, 57585, 57629, 59516, 60025, 62380, 74728, 76429, 76985, 78309, 80656, 80954, 81834, 81886, 81995, 82055, 82156, 82247, 82804, 83202, 83742, 83834, 84531, 84971, 88667, 88872, 92159, 94973, 96204, 96290, 96392, 97478, 98130, 98381, 98466, 98885, 99220, 99411, 99539, 99647, 99667, 99746, 100137, 100234, 100334, 100702, 100929, 101103, 101319, 101365, 101402, 101596, 102180, 103160, 104388, 105039, 108737, 109250, 110025, 110449, 111396, 111600, 111829, 111913, 112591, 112813, 113438, 113480, 113943, 114030, 114676, 115240, 127011]
 
Assigning carbon spectrum...
 
Plotting carbon spectrum...
 
Assigning proton spectrum...
 
Plotting proton spectrum...
 
Calculating DP4 probabilities...
none
No stats model provided, using default
number of c protons = 10
number of c carbons = 6
number of e protons = 10
number of e carbons = 6
IP4_1
 
Solvent = chloroform
Force Field = mmff
 
DFT NMR Functional = mPW1PW91
DFT NMR Basis = 6-311g(d)
 
Number of isomers = 2
Number of conformers for isomer 1 = 3
Number of conformers for isomer 2 = 5
 
Assigned C shifts for isomer 1:
label, calc, corrected, exp, error
C7      18.93  15.82  16.23   0.41
C2      43.81  39.16  39.34   0.18
C4      53.76  48.50  47.15  -1.35
C3      81.42  74.44  75.43   0.99
C1     136.44 126.05 125.26  -0.79
C5     143.34 132.52 133.08   0.56
 
Assigned C shifts for isomer 2:
label, calc, corrected, exp, error
C7      14.73  15.72  16.23   0.51
C2      45.14  43.41  39.34  -4.07
C4      47.43  45.50  47.15   1.65
C3      76.56  72.03  75.43   3.40
C1     137.20 127.26 125.26  -2.00
C5     143.03 132.57 133.08   0.51
 
Assigned H shifts for isomer 1:
label, calc, corrected, exp, error
H15      0.60   0.71   0.95   0.24
H14      0.86   0.95   1.66   0.71
H17      1.15   1.22   0.95  -0.27
H16      1.28   1.34   0.95  -0.39
H10      2.29   2.27   2.18  -0.10
H9       2.61   2.58   2.55  -0.02
H8       2.75   2.71   2.55  -0.15
H11      4.05   3.91   3.91  -0.01
H12      5.76   5.50   5.54   0.04
H13      5.87   5.61   5.54  -0.06
 
Assigned H shifts for isomer 2:
label, calc, corrected, exp, error
H14      0.69   0.73   1.66   0.93
H16      0.98   1.00   0.95  -0.05
H17      1.09   1.10   0.95  -0.15
H15      1.23   1.23   0.95  -0.28
H8       2.42   2.35   2.18  -0.17
H9       2.70   2.62   2.55  -0.07
H10      2.72   2.63   2.55  -0.08
H11      4.31   4.13   3.91  -0.23
H13      5.71   5.45   5.54   0.10
H12      5.81   5.54   5.54   0.00
 
Results of DP4 using Proton:
Isomer 1: 97.2%
Isomer 2:  2.8%
 
Results of DP4 using Carbon:
Isomer 1: 99.3%
Isomer 2:  0.7%
 
Results of DP4:
Isomer 1: 100.0%
Isomer 2:  0.0%
 
PyDP4 process completed successfully.

可以看到,本次DP4-AI工作流对化合物进行了差向异构体枚举,得到两个结构isomer 1与Isomer 2; 然后分别对两个异构体进行构象搜索,异构体1、2分别得到3与5个构象;分别对每个构象进行NMR计算、根据构象能得到计算的NMR数据;并提取实验NMR数据;然后进行DP4计算,分别给出了NMR实验数据归属为两个异构体的DP4概率。

最重要的是,在归属的过程中,我们没有事先对1H与13C-NMR实验数据进行预处理与归属,这全部由DP4-AI自动动进行。比起需要花大量的时间编写NMR描述数据的常规方法DP4计算,DP4-AI的自动NMR数据处理为化学科研人员节省了大量的宝贵时间。

除了给出DP4概率之外,计算与实验图谱的归属也生成在一个新的目录Graphs里,每个异构体分别一张13C与1H-NMR图谱, 以矢量图方式给出。图11与12是其中一个异构体的13C与1H-NMR图谱。

DP4-AI与Pairwise AA的正确预测率

Figure 11. 13C-NMR实验与计算图谱及其归属(矢量图,可缩放)

DP4-AI与Pairwise AA的正确预测率

Figure 11. 1H-NMR实验与计算图谱及其归属(矢量图,可缩放)

积分也是自动完成,生成的矢量图里已经指认好了计算峰与实验峰,通过颜色进行对映,溶剂峰的着色也区别其它峰。结合给出的各个异构体的SDF文件,可以很方便的将化合物编号与图谱编号进行比对解读。

除了给出图形格式的NMR数据,还新生成了一个目录Pickles, 该目录将将实验NMR数据以数据文件的格式保存,可以用Pickle模块读取。此外,还新生成文件ANPN-D-4-60_1NMR.dp4,包含了dp4计算过程的信息信息。新生成的两个目录与一个文件结构如下:

1
2
3
4
5
6
7
8
9
10
11
Graphs
└── IP4_1
    ├── Carbon_1.svg
    ├── Carbon_2.svg
    ├── Proton_1.svg
    └── Proton_2.svg
Pickles
└── IP4_1
    ├── carbondata
    └── protondata
IP4_1NMR.dp4

注意:如果你有第二个异构体的NMR数据,重新进行DP4计算此时,需要将Graphs与Pickles更名再计算DP4,并且IP4_1NMR.dp4会被重写。

手工计算

根据博文《DP4概率的计算》所描述的方法,Isomer_1/2的学生氏t-分布与Gaussin概率分布计算的DP4概率分别为98.97%/99.46%与1.03%/0.44%。这与DP4-AI自动计算的DP4概率99.3%与0.7%基本一致。

注意事项

DP4-AI自动读取实验数据并进行归属与积分并非万无一失。实践中发现,很多因素会导致自动积分错误与归属错误。比如有杂质峰的时候会出现积分错误和/或归属错误,虽然此类错误在化学工作者看来显而易见。因此,强烈建议:在采用DP4之前,对生成的图谱与归属进行复核。

DP4计算的布署与培训

部署与培训服务,请联系我们。