paccmann_predictor / Git / Diff of /README.md

Models:

Amanda-D/

paccmann_predictor

Downloads: 1

Diff of /README.md [992277] .. [203523]

Switch to unified view


[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
[![Build Status](https://github.com/PaccMann/paccmann_predictor/actions/workflows/build.yml/badge.svg)](https://github.com/PaccMann/paccmann_predictor/actions/workflows/build.yml)

# paccmann_predictor

Drug interaction prediction with PaccMann.

`paccmann_predictor` is a package for drug interaction prediction, with examples of 
anticancer drug sensitivity prediction and drug target affinity prediction. Please see our papers:

- [_Toward explainable anticancer compound sensitivity prediction via multimodal attention-based convolutional encoders_](https://doi.org/10.1021/acs.molpharmaceut.9b00520) (*Molecular Pharmaceutics*, 2019). This is the original paper on IC50 prediction using drug properties and tissue-specific cell line information (gene expression profiles). While the original code was written in `tensorflow` and is available [here](https://github.com/drugilsberg/paccmann), this is the `pytorch` implementation of the best PaccMann architecture (multiscale convolutional encoder).


**PaccMann for affinity prediction:**
- [Data-driven molecular design for discovery and synthesis of novel ligands: a case study on SARS-CoV-2](https://iopscience.iop.org/article/10.1088/2632-2153/abe808) (_Machine Learning: Science and Technology_, 2021). In there, we propose a slightly modified version to predict drug-target binding affinities based on protein sequences and SMILES

![Graphical abstract](https://github.com/PaccMann/paccmann_predictor/blob/master/assets/paccmann.png?raw=true "Graphical abstract")

## Installation
The library itself has few dependencies (see [setup.py](setup.py)) with loose requirements. 
First, set up the environment as follows:
```sh
conda env create -f examples/IC50/conda.yml
conda activate paccmann_predictor
pip install -e .
```


## Evaluate pretrained drug sensitivty model on your own data
First, please consider using our public [PaccMann webservice](https://ibm.biz/paccmann-aas) as described in the [NAR paper](https://academic.oup.com/nar/article/48/W1/W502/5836770).

To use our pretrained model, please download the model from: https://ibm.biz/paccmann-data (just download `models/single_pytorch_model`).
For example, assuming that you:
1. Set up your conda environment as described above;
2. Downloaded the model linked above in a directory called `single_pytorch_model` and
3. Downloaded the data from https://ibm.box.com/v/paccmann-pytoda-data in folders `data` and `splitted_data`;
then, the following command should work:
```console
(paccmann_predictor) $ python examples/IC50/test_paccmann.py \
splitted_data/gdsc_cell_line_ic50_test_fraction_0.1_id_997_seed_42.csv \
data/gene_expression/gdsc-rnaseq_gene-expression.csv \
data/smiles/gdsc.smi \
data/2128_genes.pkl \
single_pytorch_model/smiles_language \
single_pytorch_model/weights/best_mse_paccmann_v2.pt \
results \
single_pytorch_model/model_params.json
```
*NOTE*: If you bring your own data, please make sure to provide the omic data for the 2128 genes specified in `data/2128_genes.pkl`. Your omic data (here it is `data/gene_expression/gdsc-rnaseq_gene-expression.csv`) can contain more columns and it does not need to follow the order of the pickled gene list. But please dont change this pickle file. Also note that this is PaccMannV2 which is slightly improved compared to the paper version (context attention on both modalities).

## Finetuning on your own data
You can also **finetune** our pretrained model on your data instead of training a model from scratch. For that, please follow the instruction below for training on scratch and just set:
- `model_path` --> directory where the `single_pytorch_model` is stored
- `training_name` --> this should be `single_pytorch_model`
- `params_filepath` --> `base_path/single_pytorch_model/model_params.json`


## Training a model from scratch
To run the example training script we provide environment files under `examples/IC50/`.
In the `examples` directory is a training script [train_paccmann.py](./examples/IC50/train_paccmann.py) that makes use
of `paccmann_predictor`.

```console
(paccmann_predictor) $ python examples/IC50/train_paccmann.py -h
usage: train_paccmann.py [-h]
                         train_sensitivity_filepath test_sensitivity_filepath
                         gep_filepath smi_filepath gene_filepath
                         smiles_language_filepath model_path params_filepath
                         training_name

positional arguments:
  train_sensitivity_filepath
                        Path to the drug sensitivity (IC50) data.
  test_sensitivity_filepath
                        Path to the drug sensitivity (IC50) data.
  gep_filepath          Path to the gene expression profile data.
  smi_filepath          Path to the SMILES data.
  gene_filepath         Path to a pickle object containing list of genes.
  smiles_language_filepath
                        Path to a pickle object a SMILES language object.
  model_path            Directory where the model will be stored.
  params_filepath       Path to the parameter file.
  training_name         Name for the training.

optional arguments:
  -h, --help            show this help message and exit
```

`params_filepath` could point to [examples/IC50/example_params.json](examples/IC50/example_params.json), examples for other files can be downloaded from [here](https://ibm.box.com/v/paccmann-pytoda-data).

## References

If you use `paccmann_predictor` in your projects, please cite the following:

```bib
@article{manica2019paccmann,
  title={Toward explainable anticancer compound sensitivity prediction via multimodal attention-based convolutional encoders},
  author={Manica, Matteo and Oskooei, Ali and Born, Jannis and Subramanian, Vigneshwari and S{\'a}ez-Rodr{\'\i}guez, Julio and Mart{\'\i}nez, Mar{\'\i}a Rodr{\'\i}guez},
  journal={Molecular pharmaceutics},
  volume={16},
  number={12},
  pages={4797--4806},
  year={2019},
  publisher={ACS Publications},
  doi = {10.1021/acs.molpharmaceut.9b00520},
  note = {PMID: 31618586}
}

@article{born2021datadriven,
  author = {Born, Jannis and Manica, Matteo and Cadow, Joris and Markert, Greta and Mill, Nil Adell and Filipavicius, Modestas and Janakarajan, Nikita and Cardinale, Antonio and Laino, Teodoro and {Rodr{\'{i}}guez Mart{\'{i}}nez}, Mar{\'{i}}a},
  doi = {10.1088/2632-2153/abe808},
  issn = {2632-2153},
  journal = {Machine Learning: Science and Technology},
  number = {2},
  pages = {025024},
  title = {{Data-driven molecular design for discovery and synthesis of novel ligands: a case study on SARS-CoV-2}},
  url = {https://iopscience.iop.org/article/10.1088/2632-2153/abe808},
  volume = {2},
  year = {2021}
}
```

	a/README.md		b/README.md
1	[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)	1	[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
2	[![Build Status](https://github.com/PaccMann/paccmann_predictor/actions/workflows/build.yml/badge.svg)](https://github.com/PaccMann/paccmann_predictor/actions/workflows/build.yml)	2	[![Build Status](https://github.com/PaccMann/paccmann_predictor/actions/workflows/build.yml/badge.svg)](https://github.com/PaccMann/paccmann_predictor/actions/workflows/build.yml)
3		3
4	# paccmann_predictor	4	# paccmann_predictor
5		5
6	Drug interaction prediction with PaccMann.	6	Drug interaction prediction with PaccMann.
7		7
8	`paccmann_predictor` is a package for drug interaction prediction, with examples of	8	`paccmann_predictor` is a package for drug interaction prediction, with examples of
9	anticancer drug sensitivity prediction and drug target affinity prediction. Please see our papers:	9	anticancer drug sensitivity prediction and drug target affinity prediction. Please see our papers:
10		10
11	- [_Toward explainable anticancer compound sensitivity prediction via multimodal attention-based convolutional encoders_](https://doi.org/10.1021/acs.molpharmaceut.9b00520) (Molecular Pharmaceutics, 2019). This is the original paper on IC50 prediction using drug properties and tissue-specific cell line information (gene expression profiles). While the original code was written in `tensorflow` and is available [here](https://github.com/drugilsberg/paccmann), this is the `pytorch` implementation of the best PaccMann architecture (multiscale convolutional encoder).	11	- [_Toward explainable anticancer compound sensitivity prediction via multimodal attention-based convolutional encoders_](https://doi.org/10.1021/acs.molpharmaceut.9b00520) (Molecular Pharmaceutics, 2019). This is the original paper on IC50 prediction using drug properties and tissue-specific cell line information (gene expression profiles). While the original code was written in `tensorflow` and is available [here](https://github.com/drugilsberg/paccmann), this is the `pytorch` implementation of the best PaccMann architecture (multiscale convolutional encoder).
12		12
13		13
14	PaccMann for affinity prediction:	14	PaccMann for affinity prediction:
15	- [Data-driven molecular design for discovery and synthesis of novel ligands: a case study on SARS-CoV-2](https://iopscience.iop.org/article/10.1088/2632-2153/abe808) (_Machine Learning: Science and Technology_, 2021). In there, we propose a slightly modified version to predict drug-target binding affinities based on protein sequences and SMILES	15	- [Data-driven molecular design for discovery and synthesis of novel ligands: a case study on SARS-CoV-2](https://iopscience.iop.org/article/10.1088/2632-2153/abe808) (_Machine Learning: Science and Technology_, 2021). In there, we propose a slightly modified version to predict drug-target binding affinities based on protein sequences and SMILES
16		16
17	![Graphical abstract](https://github.com/PaccMann/paccmann_predictor/blob/master/assets/paccmann.png "Graphical abstract")	17	![Graphical abstract](https://github.com/PaccMann/paccmann_predictor/blob/master/assets/paccmann.png?raw=true "Graphical abstract")
18		18
19	## Installation	19	## Installation
20	The library itself has few dependencies (see [setup.py](setup.py)) with loose requirements.	20	The library itself has few dependencies (see [setup.py](setup.py)) with loose requirements.
21	First, set up the environment as follows:	21	First, set up the environment as follows:
22	```sh	22	```sh
23	conda env create -f examples/IC50/conda.yml	23	conda env create -f examples/IC50/conda.yml
24	conda activate paccmann_predictor	24	conda activate paccmann_predictor
25	pip install -e .	25	pip install -e .
26	```	26	```
27		27
28		28
29	## Evaluate pretrained drug sensitivty model on your own data	29	## Evaluate pretrained drug sensitivty model on your own data
30	First, please consider using our public [PaccMann webservice](https://ibm.biz/paccmann-aas) as described in the [NAR paper](https://academic.oup.com/nar/article/48/W1/W502/5836770).	30	First, please consider using our public [PaccMann webservice](https://ibm.biz/paccmann-aas) as described in the [NAR paper](https://academic.oup.com/nar/article/48/W1/W502/5836770).
31		31
32	To use our pretrained model, please download the model from: https://ibm.biz/paccmann-data (just download `models/single_pytorch_model`).	32	To use our pretrained model, please download the model from: https://ibm.biz/paccmann-data (just download `models/single_pytorch_model`).
33	For example, assuming that you:	33	For example, assuming that you:
34	1. Set up your conda environment as described above;	34	1. Set up your conda environment as described above;
35	2. Downloaded the model linked above in a directory called `single_pytorch_model` and	35	2. Downloaded the model linked above in a directory called `single_pytorch_model` and
36	3. Downloaded the data from https://ibm.box.com/v/paccmann-pytoda-data in folders `data` and `splitted_data`;	36	3. Downloaded the data from https://ibm.box.com/v/paccmann-pytoda-data in folders `data` and `splitted_data`;
37	then, the following command should work:	37	then, the following command should work:
38	```console	38	```console
39	(paccmann_predictor) $ python examples/IC50/test_paccmann.py \	39	(paccmann_predictor) $ python examples/IC50/test_paccmann.py \
40	splitted_data/gdsc_cell_line_ic50_test_fraction_0.1_id_997_seed_42.csv \	40	splitted_data/gdsc_cell_line_ic50_test_fraction_0.1_id_997_seed_42.csv \
41	data/gene_expression/gdsc-rnaseq_gene-expression.csv \	41	data/gene_expression/gdsc-rnaseq_gene-expression.csv \
42	data/smiles/gdsc.smi \	42	data/smiles/gdsc.smi \
43	data/2128_genes.pkl \	43	data/2128_genes.pkl \
44	single_pytorch_model/smiles_language \	44	single_pytorch_model/smiles_language \
45	single_pytorch_model/weights/best_mse_paccmann_v2.pt \	45	single_pytorch_model/weights/best_mse_paccmann_v2.pt \
46	results \	46	results \
47	single_pytorch_model/model_params.json	47	single_pytorch_model/model_params.json
48	```	48	```
49	NOTE: If you bring your own data, please make sure to provide the omic data for the 2128 genes specified in `data/2128_genes.pkl`. Your omic data (here it is `data/gene_expression/gdsc-rnaseq_gene-expression.csv`) can contain more columns and it does not need to follow the order of the pickled gene list. But please dont change this pickle file. Also note that this is PaccMannV2 which is slightly improved compared to the paper version (context attention on both modalities).	49	NOTE: If you bring your own data, please make sure to provide the omic data for the 2128 genes specified in `data/2128_genes.pkl`. Your omic data (here it is `data/gene_expression/gdsc-rnaseq_gene-expression.csv`) can contain more columns and it does not need to follow the order of the pickled gene list. But please dont change this pickle file. Also note that this is PaccMannV2 which is slightly improved compared to the paper version (context attention on both modalities).
50		50
51	## Finetuning on your own data	51	## Finetuning on your own data
52	You can also finetune our pretrained model on your data instead of training a model from scratch. For that, please follow the instruction below for training on scratch and just set:	52	You can also finetune our pretrained model on your data instead of training a model from scratch. For that, please follow the instruction below for training on scratch and just set:
53	- `model_path` --> directory where the `single_pytorch_model` is stored	53	- `model_path` --> directory where the `single_pytorch_model` is stored
54	- `training_name` --> this should be `single_pytorch_model`	54	- `training_name` --> this should be `single_pytorch_model`
55	- `params_filepath` --> `base_path/single_pytorch_model/model_params.json`	55	- `params_filepath` --> `base_path/single_pytorch_model/model_params.json`
56		56
57		57
58	## Training a model from scratch	58	## Training a model from scratch
59	To run the example training script we provide environment files under `examples/IC50/`.	59	To run the example training script we provide environment files under `examples/IC50/`.
60	In the `examples` directory is a training script [train_paccmann.py](./examples/IC50/train_paccmann.py) that makes use	60	In the `examples` directory is a training script [train_paccmann.py](./examples/IC50/train_paccmann.py) that makes use
61	of `paccmann_predictor`.	61	of `paccmann_predictor`.
62		62
63	```console	63	```console
64	(paccmann_predictor) $ python examples/IC50/train_paccmann.py -h	64	(paccmann_predictor) $ python examples/IC50/train_paccmann.py -h
65	usage: train_paccmann.py [-h]	65	usage: train_paccmann.py [-h]
66	train_sensitivity_filepath test_sensitivity_filepath	66	train_sensitivity_filepath test_sensitivity_filepath
67	gep_filepath smi_filepath gene_filepath	67	gep_filepath smi_filepath gene_filepath
68	smiles_language_filepath model_path params_filepath	68	smiles_language_filepath model_path params_filepath
69	training_name	69	training_name
70		70
71	positional arguments:	71	positional arguments:
72	train_sensitivity_filepath	72	train_sensitivity_filepath
73	Path to the drug sensitivity (IC50) data.	73	Path to the drug sensitivity (IC50) data.
74	test_sensitivity_filepath	74	test_sensitivity_filepath
75	Path to the drug sensitivity (IC50) data.	75	Path to the drug sensitivity (IC50) data.
76	gep_filepath Path to the gene expression profile data.	76	gep_filepath Path to the gene expression profile data.
77	smi_filepath Path to the SMILES data.	77	smi_filepath Path to the SMILES data.
78	gene_filepath Path to a pickle object containing list of genes.	78	gene_filepath Path to a pickle object containing list of genes.
79	smiles_language_filepath	79	smiles_language_filepath
80	Path to a pickle object a SMILES language object.	80	Path to a pickle object a SMILES language object.
81	model_path Directory where the model will be stored.	81	model_path Directory where the model will be stored.
82	params_filepath Path to the parameter file.	82	params_filepath Path to the parameter file.
83	training_name Name for the training.	83	training_name Name for the training.
84		84
85	optional arguments:	85	optional arguments:
86	-h, --help show this help message and exit	86	-h, --help show this help message and exit
87	```	87	```
88		88
89	`params_filepath` could point to [examples/IC50/example_params.json](examples/IC50/example_params.json), examples for other files can be downloaded from [here](https://ibm.box.com/v/paccmann-pytoda-data).	89	`params_filepath` could point to [examples/IC50/example_params.json](examples/IC50/example_params.json), examples for other files can be downloaded from [here](https://ibm.box.com/v/paccmann-pytoda-data).
90		90
91	## References	91	## References
92		92
93	If you use `paccmann_predictor` in your projects, please cite the following:	93	If you use `paccmann_predictor` in your projects, please cite the following:
94		94
95	```bib	95	```bib
96	@article{manica2019paccmann,	96	@article{manica2019paccmann,
97	title={Toward explainable anticancer compound sensitivity prediction via multimodal attention-based convolutional encoders},	97	title={Toward explainable anticancer compound sensitivity prediction via multimodal attention-based convolutional encoders},
98	author={Manica, Matteo and Oskooei, Ali and Born, Jannis and Subramanian, Vigneshwari and S{\'a}ez-Rodr{\'\i}guez, Julio and Mart{\'\i}nez, Mar{\'\i}a Rodr{\'\i}guez},	98	author={Manica, Matteo and Oskooei, Ali and Born, Jannis and Subramanian, Vigneshwari and S{\'a}ez-Rodr{\'\i}guez, Julio and Mart{\'\i}nez, Mar{\'\i}a Rodr{\'\i}guez},
99	journal={Molecular pharmaceutics},	99	journal={Molecular pharmaceutics},
100	volume={16},	100	volume={16},
101	number={12},	101	number={12},
102	pages={4797--4806},	102	pages={4797--4806},
103	year={2019},	103	year={2019},
104	publisher={ACS Publications},	104	publisher={ACS Publications},
105	doi = {10.1021/acs.molpharmaceut.9b00520},	105	doi = {10.1021/acs.molpharmaceut.9b00520},
106	note = {PMID: 31618586}	106	note = {PMID: 31618586}
107	}	107	}
108		108
109	@article{born2021datadriven,	109	@article{born2021datadriven,
110	author = {Born, Jannis and Manica, Matteo and Cadow, Joris and Markert, Greta and Mill, Nil Adell and Filipavicius, Modestas and Janakarajan, Nikita and Cardinale, Antonio and Laino, Teodoro and {Rodr{\'{i}}guez Mart{\'{i}}nez}, Mar{\'{i}}a},	110	author = {Born, Jannis and Manica, Matteo and Cadow, Joris and Markert, Greta and Mill, Nil Adell and Filipavicius, Modestas and Janakarajan, Nikita and Cardinale, Antonio and Laino, Teodoro and {Rodr{\'{i}}guez Mart{\'{i}}nez}, Mar{\'{i}}a},
111	doi = {10.1088/2632-2153/abe808},	111	doi = {10.1088/2632-2153/abe808},
112	issn = {2632-2153},	112	issn = {2632-2153},
113	journal = {Machine Learning: Science and Technology},	113	journal = {Machine Learning: Science and Technology},
114	number = {2},	114	number = {2},
115	pages = {025024},	115	pages = {025024},
116	title = {{Data-driven molecular design for discovery and synthesis of novel ligands: a case study on SARS-CoV-2}},	116	title = {{Data-driven molecular design for discovery and synthesis of novel ligands: a case study on SARS-CoV-2}},
117	url = {https://iopscience.iop.org/article/10.1088/2632-2153/abe808},	117	url = {https://iopscience.iop.org/article/10.1088/2632-2153/abe808},
118	volume = {2},	118	volume = {2},
119	year = {2021}	119	year = {2021}
120	}	120	}
121	```	121	```