Overview¶
You need to decompress the example_data.zip file in your preferred folder, e.g., deeptminter/.
Deep learning for interaction site prediction¶
Python¶
We define several parameters required for running .
params = {
'prot_name': '3jcu',
'prot_chain': 'H',
'fasta_fp': '../../data/deeptminter/example_data/',
'msa_fp': '../../data/deeptminter/example_data/',
'phobius_fp': '../../data/deeptminter/example_data/',
'mi_fp': '../../data/deeptminter/example_data/',
'fc_fp': '../../data/deeptminter/example_data/',
'gdca_fp': '../../data/deeptminter/example_data/',
'sv_fp_feature': '../../data/deeptminter/example_data/',
'sv_suffix_feature': '.f',
'model_frozen_fpn': '../../data/deeptminter/model/tf2/frozen_graph/m1.pb',
'batch_size': 100,
'sv_fp_pred': '../../data/deeptminter/',
'sv_suffix_pred': '.m1',
}Code
Output
import deeptminter
deeptminter.predict.isite(
prot_name=params['prot_name'],
prot_chain=params['prot_chain'],
fasta_fp=params['fasta_fp'],
msa_fp=params['msa_fp'],
phobius_fp=params['phobius_fp'],
mi_fp=params['mi_fp'],
fc_fp=params['fc_fp'],
gdca_fp=params['gdca_fp'],
sv_fp_feature=params['sv_fp_feature'],
sv_suffix_feature=params['sv_suffix_feature'],
model_frozen_fpn=params['model_frozen_fpn'],
sv_fp_pred=params['sv_fp_pred'],
sv_suffix_pred=params['sv_suffix_pred'],
batch_size=100,
verbose=True,
) ____ _____ __ __ ___ _
| _ \ ___ ___ _ _|_ _| \/ |_ _|_ __ | |_ ___ _ __
| | | |/ _ \/ _ \ '_ \| | | |\/| || || '_ \| __/ _ \ '__|
| |_| | __/ __/ |_) | | | | | || || | | | || __/ |
|____/ \___|\___| .__/|_| |_| |_|___|_| |_|\__\___|_|
|_|
05/04/2025 16:00:51 logger: ===>Protein: 3jcu chain: H
05/04/2025 16:00:51 logger: ======>Features are being assembled...
05/04/2025 16:00:51 logger: ======>Finished being assembled.
05/04/2025 16:00:51 logger: ===>The model (../../data/deeptminter/model/tf2/frozen_graph/m1.pb) with frozen graphs converted from tensorflow 1.15.2 is being read...
05/04/2025 16:00:53 logger: ======>Predicted probabilities of interation sites:
0 1 2
0 1 P 0.44524422
1 2 K 0.5642183
2 3 P 0.57570034
...
57 58 S 0.53011745
58 59 M 0.3377973
05/04/2025 16:00:53 logger: ======>Predictions are saved to ../../data/deeptminter/3jcuH.m1CLI¶
DeepTMInter can also be used in shell. To know how to use, please type
deeptminter -hIt shows the usage of different parameters.
-pn, --prot_name, Name of the protein to be processed.
-pc, --prot_chain, Chain ID of the protein (e.g., A, B, etc.).
-fa, --fasta_fp, File path to the input FASTA sequence.
-msafp, --msa_fp, File path to the multiple sequence alignment (MSA).
-phobfp, --phobius_fp, File path to the Phobius output file.
-mifp, --mi_fp, File path to the mutual information (MI) data.
-fcfp, --fc_fp, File path to the FreeContact (FC) data.
-gdcafp, --gdca_fp, File path to the direct coupling analysis (DCA) output.
-m, --model_frozen_fpn, Path to the frozen model (.pb) file for prediction.
-sv_fp_f, --sv_fp_feature, Path to the feature vector for interaction site prediction.
-sv_suf_f, --sv_suffix_feature, Suffix for feature files used in interaction site input.
-sv_fp_p, --sv_fp_pred, Output file path for predicted results.
-sv_suf_p, --sv_suffix_pred, Suffix for predicted output files.
-bs, --batch_size, Batch size used during prediction (default: 100).
-vb, --verbose, Whether to print detailed logs during processing (default: True).You can run it using the following code.
Command
Output
deeptminter -pn 3jcu -pc H -fa ./data/deeptminter/example_data/ -msafp ./data/deeptminter/example_data/ -phobfp ./data/deeptminter/example_data/ -mifp ./data/deeptminter/example_data/ -fcfp ./data/deeptminter/example_data/ -gdcafp ./data/deeptminter/example_data/ -m ./data/deeptminter/model/tf2/frozen_graph/m1.pb -sv_fp_f ./data/deeptminter/example_data/ -sv_suf_f .f -sv_fp_p ./data/deeptminter/ -sv_suf_p .m1 -bs 100 ____ _____ __ __ ___ _
| _ \ ___ ___ _ _|_ _| \/ |_ _|_ __ | |_ ___ _ __
| | | |/ _ \/ _ \ '_ \| | | |\/| || || '_ \| __/ _ \ '__|
| |_| | __/ __/ |_) | | | | | || || | | | || __/ |
|____/ \___|\___| .__/|_| |_| |_|___|_| |_|\__\___|_|
|_|
05/04/2025 16:11:18 logger: ===>Protein: 3jcu chain: H
05/04/2025 16:11:18 logger: ======>Features are being assembled...
05/04/2025 16:11:19 logger: ======>Finished being assembled.
05/04/2025 16:11:19 logger: ===>The model (./data/deeptminter/model/tf2/frozen_graph/m1.pb) with frozen graphs converted from tensorflow 1.15.2 is being read...
05/04/2025 16:11:20 logger: ===>The model is read in, with
05/04/2025 16:11:20 logger: ======>Predicted probabilities of interation sites:
0 1 2
0 1 P 0.44524422
1 2 K 0.5642183
2 3 P 0.57570034
...
57 58 S 0.53011745
58 59 M 0.3377973
05/04/2025 16:11:20 logger: ======>Predictions are saved to ./data/deeptminter/3jcuH.m1Assembled prediction¶
If you’d like to explore if better predictions can be gained from assembled methods, you can generate models by choosing to run the running code.
Python¶
First, we define parameters for input and output.
params_stacking = {
'prot_name': '3jcu',
'prot_chain': 'H',
'region': 'combined',
'fasta_fp': '../../data/deeptminter/example_data/',
'phobius_fp': '../../data/deeptminter/example_data/',
'isite_fp': '../../data/deeptminter/',
'model_fpn': '../../data/deeptminter/model/stacking.model',
'sv_fp_stacking_input': '../../data/deeptminter/',
'sv_fp': '../../data/deeptminter/',
'sv_suffix': '.deeptminter'
}Then, we performed predictions.
Command
Output
import deeptminter
deeptminter.predict.stacking(
prot_name=params_stacking['prot_name'],
prot_chain=params_stacking['prot_chain'],
region=params_stacking['region'],
fasta_fp=params_stacking['fasta_fp'],
phobius_fp=params_stacking['phobius_fp'],
isite_fp=params_stacking['isite_fp'],
sv_fp_stacking_input=params_stacking['sv_fp_stacking_input'],
model_fpn=params_stacking['model_fpn'],
sv_fp=params_stacking['sv_fp'],
sv_suffix=params_stacking['sv_suffix'],
) ____ _____ __ __ ___ _
| _ \ ___ ___ _ _|_ _| \/ |_ _|_ __ | |_ ___ _ __
| | | |/ _ \/ _ \ '_ \| | | |\/| || || '_ \| __/ _ \ '__|
| |_| | __/ __/ |_) | | | | | || || | | | || __/ |
|____/ \___|\___| .__/|_| |_| |_|___|_| |_|\__\___|_|
|_|
05/04/2025 16:33:23 logger: ===>Protein: 3jcu chain: H
05/04/2025 16:33:23 logger: Predictions are:
0 1 2
0 29 L 0.6798641623872085
1 30 M 0.6557471888350399
2 31 G 0.6949745997232504
...
19 48 Y 0.7044071371080728
20 49 N 0.705696071814507
05/04/2025 16:33:23 logger: Finished!CLI¶
DeepTMInter can also be used in shell. To know how to use, please type
Command
Output
deeptminter_assemble -pn 3jcu -pc H -fa ./data/deeptminter/example_data/ -phobfp ./data/deeptminter/example_data/ -m ./data/deeptminter/model/stacking.model -ifp ./data/deeptminter/ -r combined -sv_fp_s ./data/deeptminter/ -sv_fp ./data/deeptminter/ -sv_suf .deeptminter ____ _____ __ __ ___ _
| _ \ ___ ___ _ _|_ _| \/ |_ _|_ __ | |_ ___ _ __
| | | |/ _ \/ _ \ '_ \| | | |\/| || || '_ \| __/ _ \ '__|
| |_| | __/ __/ |_) | | | | | || || | | | || __/ |
|____/ \___|\___| .__/|_| |_| |_|___|_| |_|\__\___|_|
|_|
05/04/2025 16:42:07 logger: ===>Protein: 3jcu chain: H
05/04/2025 16:42:08 logger: Predictions are:
0 1 2
0 29 L 0.6798641623872085
1 30 M 0.6557471888350399
2 31 G 0.6949745997232504
3 32 V 0.6678435513228781
4 33 A 0.4116563950495269
5 34 M 0.6756479650668131
6 35 A 0.3599768716465631
7 36 L 0.5715904491452858
8 37 F 0.7324822410039901
9 38 A 0.4415552709397562
10 39 V 0.6785871628435665
11 40 F 0.5630136567015621
12 41 L 0.6344924489703325
13 42 S 0.7063788778070768
14 43 I 0.6599980449464243
15 44 I 0.6579266981626475
16 45 L 0.7058936221993016
17 46 E 0.639440562706253
18 47 I 0.6948413389284585
19 48 Y 0.7044071371080728
20 49 N 0.705696071814507
05/04/2025 16:42:08 logger: Finished!