easyaskcos 2.1.0

Installation

Installing this package depends on a working installation of Python with pip. To install the package, run:

pip install git+ssh://git@github.com:syngenta/EasyAskcos.git

For Development

When working on the development of this package, the developer wants to work directly on the source code while still using the packaged installation. For that, run:

git clone git@github.com:syngenta/EasyAskcos.git
pip install -e easyaskcos/[dev]

Configuration

This package requires some configuration parameters to work, including some secretes and settings. After installation, and before the first usage, the useR should run the following command

easyaskcos_configure

this command generates the user_home/easyaskcos directory and places into it two files:

1. settings.toml populated with defaults settings. The user can review and modify these values if necessary

2. .secrets.toml containing the keys for the necessary secrets. The user must replace the placeholders with the correct values

Usage: ASKCOS V1

Let’s initialize the wrapper.

from easyaskcos.wrapper import AskcosWrapper

wrapper = AskcosWrapper()

It will connect to the url provided in the settings file, however it is possible to define a different base url by indicating it during the initialization of the wrapper

from easyaskcos.wrapper import AskcosWrapper

wrapper = AskcosWrapper(base_url="https://some.server")

Now it is possible to access various endpoints provided by the Askcos API in a simple and straightforward manner.

Perform prediction for retrosynthetic trees

response = wrapper.predict_retrosynthetic_tree("CCC(O)=O")
# to check the status of the response
prediction_id = response["prediction_id"]
results = wrapper.get_results(prediction_id)
print(results["status"])
# upon "SUCCESS", the results can be inspected
retrosynthetic_trees = results["output"]

To inspect the available template sets for the retrosynthetic predictions

sets = wrapper.list_template_sets()

Each string in the returned list can be assigned to the “template_set” argument of the “predict_retrosynthetic_tree” method; by default the “reaxys” template is used.

response = wrapper.predict_retrosynthetic_tree("CCC(O)=O", template_set="pistachio")

Perform prediction for chemical reactions

response = wrapper.predict_forward("CC(=O)c1cc(Cl)ccc1Oc1ccc(Cl)cc1Cl.O=C(OO)c1cccc(Cl)c1")
# to check the status of the response
prediction_id = response["prediction_id"]
results = wrapper.get_results(prediction_id)
print(results["status"])
# upon "SUCCESS", the results can be inspected
reactions = results["output"]

Usage: ASKCOS V2

With the release of the new ASKCOS V2 platform, that soon will completely replace the previous version, we included in this package also the SDK for the new version. The usage remain overall very similar to how it is for version 1, starting with the initialization of the wrapper:

from easyaskcos.wrapper import AskcosWrapperV2

wrapper = AskcosWrapperV2()

Also in this case it will automatically connect to the url provided in the settings file, but it is always possible to specify a different base url as follows:

from easyaskcos.wrapper import AskcosWrapperV2

wrapper = AskcosWrapperV2(base_url="https://some.server")

Some endpoints of ASKCOS V2 require authentication. In order to get the token necessary to access these endpoints, you need to define your username and password for the ASKCOS V2 platform. This can be done either by adding this information in the .secrets.yaml file created by the initial configuration, or by passing these arguments while instantiating the wrapper.

from easyaskcos.wrapper import AskcosWrapperV2

wrapper = AskcosWrapperV2(user_name="my_username",
                          password="my_password")

For example, the endpoint for searching retrosynthetic trees for a given target molecule in an asynchronous way, requires the authentication. If you included the necessary credentials in the settings or while instantiating the wrapper, EasyAskcos will take care of handling the authentication for you:

from easyaskcos.wrapper import AskcosWrapperV2

wrapper = AskcosWrapperV2(user_name="my_username",
                          password="my_password")
# Retrieve the task ID associated with your prediction
task_id = wrapper.predict_retrosynthetic_tree_async("CCC(O)=O")
# Retrieve the results
results = wrapper.get_results(task_id)
print(results["status"])
# upon "SUCCESS", the results can be inspected
reactions = results["response"]["output"]

In the new version of ASKCOS, There is also an endpoint for retrosynthetic predictions that works synchronously, which does not require authentication. So you will need to just run:

from easyaskcos.wrapper import AskcosWrapperV2

results = wrapper.predict_retrosynthetic_tree("CCC(O)=O")

The endpoints for the other types of predictions (forward, impurities and reaction conditions) are still asynchronous:

task_id = wrapper.predict_forward([
                "CS(=N)(=O)Cc1cccc(Br)c1.Nc1cc(Br)ccn1",
                "CCO.CC(=O)O"
              ],)

results = wrapper.get_results(task_id)
print(results["status"])
# upon "SUCCESS", the results can be inspected
reactions = results["response"]["output"]