Reinvent: RL

AstraZeneca |

Reinvent: RL

Drug Design

Run
About

Step 1: Upload your data

Upload Prior/Agent File

Drag your file(s) or upload

Your file can be in the following formats:prior, new, agent

Prior/Agent file produced by (Reinvent: Transfer Learning) app. You can directly use its output here as an input.

Don’t have a file?

Use our demo data to run

Use Demo Data

Upload QSAR Model File

Drag your file(s) or upload

Your file can be in the following formats:pkl

Pkl file produced by Reinvent: Train QSAR Models for Target Protein. Its output can be used in this app as an input.

Don’t have a file?

Use our demo data to run

Use Demo Data

Upload Inception SMILES File (Optional)

Drag your file(s) or upload

Your file can be in the following formats:csv

Please provide SMILES in CSV format. SMILES should be in the first column without a header. The purpose of providing these SMILES in the RL process is to follow in the footsteps of well-scored small molecules previously identified.

Don’t have a file?

Use our demo data to run

Use Demo Data

Upload Matching Substructure SMARTS (Optional)

Drag your file(s) or upload

Your file can be in the following formats:csv

Please provide SMARTS in CSV format. SMARTS should be in the first column without a header. They can be used to enforce the match to a given substructure. It penalizes the total score if desired substructures are not seen in the generated compound. It produces a score of either 1 or 0.5 in the result depending on whether the desired scaffold is present or not.

Don’t have a file?

Use our demo data to run

Use Demo Data

Upload Custom Alerts SMARTS (Optional)

Drag your file(s) or upload

Your file can be in the following formats:csv

Please provide SMARTS in CSV format. SMARTS should be in the first column without a header. These SMARTS can be used enforce to NOT match a given substructure. Those who want aiming novelty and avoid certain molecular substructures can provide their SMARTS structures here.

Don’t have a file?

Use our demo data to run

Use Demo Data

Step 2: Set Parameters

Epoch

125

1000

Learning Rate

0.0001

0.9000

Scikit-Learn Model

Regression

The reference notebook of this app is "Reinforcement Learning Demo" from the ReinventCommunity repository. The purpose of this app decides which molecules are "good" or "bad" by using RL. Generally, the generative model that was created before must be directed to a relevant region of chemical space that involve compounds of interest. To achieve this, a Reinforcement Learning methodology is used with several scoring functions defined by the user. During the RL, molecules are generated by the agent and receive a score between 0 and 1. For more information, please see the tutorial page of Reinvent Apps.

Example use case: Generate molecules with RL setup.

Limitations:
-Some of the parameters are kept as default and not shown here (they will be added next versions)
-Currently, only the CPU version is available

Technology: Reinforcement Learning

Metrics: See metrics in the original demo notebook this app is derived from

Citation:

Blaschke T, Arús-Pous J, Chen H, Margreitter C, Tyrchan C, Engkvist O, et al. REINVENT 2.0 – an AI Tool for De Novo Drug Design. ChemRxiv 2020. doi:10.26434/chemrxiv.12058026.v3. This content is a preprint and has not been peer-reviewed.

Released:
Nov-21-2022

Previous Job Parameters

Your previous job parameters will show up here
so you can keep track of your jobs