ML Predict: Regression Using Fingerprints for Small Molecules

This floe predicts properties of small, drug-like molecules using a pretrained machine learning (ML) model.

It runs a TensorFlow-based fully-connected neural network regression model for prediction. This model needs to be provided by the user and can be generated using the ML Build: Regression Model with Tuner using Fingerprints for Small Molecules Floe.

This floe uses a convex box approach for domain of application predictions. The input TensorFlow dataset also contains a model agnostic system to explain the predictions on the molecule.

The user can also provide an optional TensorFlow-based probabilistic fully-connected neural network for better error bar prediction. All the models run on 2D fingerprints.

This floe runs quickly and is very inexpensive, costing about one cent for a property prediction of 10 molecules.

Outputs:

Failure Data: (a) The molecule is too large or too small, or (b) the molecule has an unknown atom.

No Confidence Data: The molecule’s property falls out of scope of training set. In this case, the model predicts with no guarantees. The explainer image has a red background.

Success Data: (a) The molecule falls within scope; the explainer has a green background, or (b) it falls at the edge of scope; the explainer has a yellow background.

Molecules outside the scope of the training set will be sent to the “No Confidence” port, as a prediction cannot be considered reliable. Specifically, the scope is defined as a range in molecular weight, atom count, polar surface area, and calculated logP from the training set molecules. These ranges are given in the Floe Report.

Inputs

Name	Description	Type
Input Small Molecule Dataset(s) to Predict Property of	The dataset(s) to read records from.	Molecule Dataset
Input TensorFlow Model	Machine learning model to predict property.	Machine Learning TensorFlow Model Dataset
Input TensorFlow Probability Model	The dataset(s) to read records from.	Machine Learning TensorFlow Probability Model Dataset

Machine Learning Model Options

Name	Description	Type
Model ID of TensorFlow Model to Use to Predict	Which model to select. Make sure this matches input model ID.	Int
Model ID of TensorFlow Probability (TFP) Model to Use to Predict*	Which model to select. Make sure this matches the input model ID.	Int
Preprocess Molecule	For every molecule, stores only largest component and adjusts ionization to neutral pH.	Bool
Apply Blockbuster Filter	Apply blockbuster filter.	Bool

Explanation and Validation

Name	Description	Type
Property Validation Field	If the dataset has a baseline, the floe provides a comparison between predictions in the Floe Report.	Float
Molecule Explainer Type	Select explainer visualization. Atom: annotate atoms only; Fragment: annotate fragments; Combined: annotate both.	List

Outputs

Name	Description	Type
Output Property	Output dataset to write to.	Dataset
Failure Property	Output dataset to write to.	Dataset