Machine Learning for Ionic Conductivity

Tutorial Created with Software Release: 2024-4
Topics: Energy Capture & Storage, Informatics and Team Collaboration
Methodology: Machine Learning
Products Used: AutoQSAR, MS Informatics, MS Maestro

Tutorial files

9.5 MB
This tutorial is written for use with a 3-button mouse with a scroll wheel.
Words found in the Glossary of Terms are shown like this: Workspacethe 3D display area in the center of the main window, where molecular structures are displayed

 

Tip: You can hover over a glossary term to display its definition. You can click on an image to expand it in the page.
Abstract:

 

In this tutorial, we will learn how to develop machine learning models to predict the experimental ionic conductivity of ionic liquids.

 

Tutorial Content

  1. Introduction

  1. Creating Projects and Importing Structures

  1. Building Machine Learning Models Using AutoQSAR

  1. Viewing the Machine Learning Models

  1. Predicting Ionic Conductivity for an Unseen Test Set

  1. Conclusion and References

  1. Glossary of Terms

1. Introduction

Quantitative Structure-Activity Relationships (QSAR) are useful modeling tools to efficiently predict material properties for a wide-range of molecules. Schrödinger’s AutoQSAR tools for generating machine learning models are easy to use, facilitating automated generation of accurate QSAR models. For practice, tutorials are available using the Materials Science (MS) Maestro suite to predict properties of small molecules, polymers, and periodic systems: Machine Learning for Materials Science, Polymer Descriptors for Machine Learning, Cheminformatics Machine Learning for Homogeneous Catalysis and Periodic Descriptors for Inorganic Solids.

The current generation of Li-ion batteries use carbonate-based electrolytes mixed with salts, such as lithium hexafluorophosphate, LiPF6, which are highly volatile and flammable. To address safety concerns arising from hazardous electrolytes, one possible solution is to replace the electrolyte with an ionic liquid (IL). Ionic liquids have good electrochemical and thermal stability, which could result in a safer battery; however, they suffer from low-to-medium ionic conductivity, which dictates how fast a battery can charge or discharge. Significant efforts have been focused on identifying ILs that have high ionic conductivities while maintaining the stability gained from using these electrolytes. In this tutorial, we will use the AutoQSAR panel in MS Maestro and an IL dataset from the NIST IL Thermo Database to create a machine learning model to predict ionic conductivity on a set of ILs at a fixed temperature of ~298.15 Kelvin (see References). A total of ~400 ILs are used to train and evaluate the machine learning models.

Figure 1. Tutorial workflow showing the 3D structures of ionic liquids in MS Maestro, which are used to train and test machine learning models with the AutoQSAR panel.

2. Creating Projects and Importing Structures

At the start of the session, change the file path to your chosen Working Directorythe location where files are saved in MS Maestro to make file navigation easier. Each session in MS Maestro begins with a default Scratch Projecta temporary project in which work is not saved, closing a scratch project removes all current work and begins a new scratch project, which is not saved. A MS Maestro project stores all your data and has a .prj extension. A project may contain numerous entries corresponding to imported structures, as well as the output of modeling-related tasks. Once a project is saved, the project is automatically saved each time a change is made.

Structures can be built in MS Maestro or can be imported using File > Import Structures (or drag-and-dropped), and are added to the Entry Lista simplified view of the Project Table that allows you to perform basic operations such as selection and inclusion and Project Tabledisplays the contents of a project and is also an interface for performing operations on selected entries, viewing properties, and organizing structures and data. The Entry Lista simplified view of the Project Table that allows you to perform basic operations such as selection and inclusion is located to the left of the Workspacethe 3D display area in the center of the main window, where molecular structures are displayed. The Project Tabledisplays the contents of a project and is also an interface for performing operations on selected entries, viewing properties, and organizing structures and data can be accessed by Ctrl+T (Cmd+T) or Window > Project Table if you would like to see an expanded view of your project data.

  1. Double-click the Materials Science icon

Figure 2-1. Change Working Directory option.

  1. Go to File > Change Working Directory
  2. Find your directory, and click Choose
  3. Pre-generated files are included for running jobs or examining output. Download the zip file here: schrodinger.com/sites/default/files/s3/release/current/Tutorials/zip/ml_ionic_conductivity.zip
  4. After downloading the zip file, unzip the contents in your Working Directorythe location where files are saved for ease of access throughout the tutorial

Figure 2-2. Save Project panel.

  1. Go to File > Save Project As
  2. Change the File name to ml_ionic_conductivity, click Save
    • The project is now named ml_ionic_conductivity.prj

Figure 2-3. Selecting the training file.

  1. Go to File > Import Structures
  2. Navigate to where you downloaded the tutorial files (presumably your working directory) and choose fixed_temp_298_K_train.csv. Click Open

Figure 2-4. Importing in the training set.

  1. Change the ENTRY TITLE Column to Index
  2. Click OK

FIgure 2-5. The ionic conductivity data set has now been imported.

352 ionic liquid structures with their respective ionic conductivity values that will be used to train AutoQSAR models have now been imported into the project

 

Note: The ionic conductivity values are associated with each entry, and can be viewed in the Project Tabledisplays the contents of a project and is also an interface for performing operations on selected entries, viewing properties, and organizing structures and data. Use the Property Tree to expose All > Canvas > Secondary > Ionic Conductivity, S/m Liquid

3. Building Machine Learning Models Using AutoQSAR

In this section, we will use the AutoQSAR panel to build machine learning models on ionic conductivity for a group of ionic liquids.

Figure 3-1. Naming the new group.

  1. Rename the group name to train_set

Figure 3-2. Opening the AutoQSAR panel.

  1. In the entry list, select(1) the atoms are chosen in the Workspace. These atoms are referred to as "the selection" or "the atom selection". Workspace operations are performed on the selected atoms. (2) The entry is chosen in the Entry List (and Project Table) and the row for the entry is highlighted. Project operations are performed on all selected entries the train_set (352) group
  2. Go to Tasks > Materials > Informatics > AutoQSAR

Figure 3-3. Setting the random training split.

  1. Set 90% for the Random training split
    • This is the percentage of data to set aside between train and test sets, where 90% of the data is used to train the model and 10% of the data is used to test the model
  2. Maintain the remaining parameters:
    • The Property to be fit is the Ionic Conductivity, S/M Liquid (canvas)
    • The Property type is numerical (as opposed to categorical)

 

More options for defining models are available in the Advanced Options dialog box, and can be referenced in the help documentation.

 

  1. Change the Job name to train_autoqsar
  2. Adjust the job settings () as needed
    • This job requires a CPU host and can be completed in about 30 minutes
  3. If you would like to run the job yourself, click Run. Otherwise, use the pre-generated train_autoqsar.qzip file from the provided tutorial files in the following section
  4. Close the AutoQSAR panel

4. Viewing the Machine Learning Models

Using the AutoQSAR panel, we can analyze the generated models.

Figure 4-1. Opening the AutoQSAR panel.

When the job is complete, note that no new entry group is added to the entry lista simplified view of the Project Table that allows you to perform basic operations such as selection and inclusion. The output can be analyzed and used for predictions in the AutoQSAR panel:

  1. Return to Tasks > Materials > Informatics > AutoQSAR
  2. For Choose task, switch to View model and make prediction
  3. To choose the Model file click Browse

Figure 4-2. Loading the .qzip file.

  1. Navigate to the Section_03 > train_autoqsar > train_autoqsar.qzip file and click Open
    • The panel will parse the .qzip file and the Model Summary section will be populated

Figure 4-3. Expanding the model report.

  1. In the Model Report section, click the + button
    • The Model Report section of the panel shows the ranking score and Q2 value (the R2 for the test set) for the best models

Several hundred 2D descriptors (molecular, topological, and feature counts) are automatically generated during the model building process and the top 10 models are retained. The models retained have performed well on the test set and have highly consistent training and test set statistics which results in a higher ranking score.

The report gives the score and Q2 value of the best-ranked models from the set of models generated in the run.

Figure 4-4. Visualizing the models.

The informative descriptors and fingerprints are used to build a large number of numeric or categorical models, where a given model is trained against a particular random subset of the input structures. The model is applied to the remaining input structures, and the accuracy of those predictions is used to arrive at an optimal number of factors for KPLS, PCR, and PLS models, and to assign an overall ranking score to the model.

 

In this case, the highest scoring models are all KPLS (Kernel partial least squares regression). For KPLS models, we can visualize atomic contributions to help understand the structural relationship to the model.

 

For each structure, each atom that contributed to a fingerprint used in building the model can be marked with a colored disk that represents the value of the contribution to the property due to that atom. The disks are blue for negative values and red for positive values. The color saturation indicates the magnitude of the contribution. Atoms that did not appear in any fingerprint are not marked with a disk.

 

  1. Click Visualize Model

Note: Ensure that all entries in the train_set group are selected before proceeding.

Figure 4-5. Viewing the conductivity.

A 2D Viewer opens in which the blue regions contribute negatively to the ionic conductivity, whereas red regions contribute positively to the ionic conductivity. Such analysis may be useful for the future design of ionic liquids.

 

Feel free to explore the various structures in the viewer.

 

 

Close the 2D Viewer panel once finished.

Figure 4-6. Opening the Report.

  1. Click Report Details to view the detailed report for the kpls_dendritic_38 model

Figure 4-7. Viewing the Report Details.

The Report Details includes the ranking scores as well as the statistical values associated with the training and test sets. In addition, each IL and its observed versus predicted value can be found here.

 

  1. To plot the data click Scatter Plot

Figure 4-8. Viewing the scatter plot.

This scatter plot visualizes the parity plot between predicted and actual ionic conductivity

  1. Close the Scatter Plot, Report Details and AutoQSAR panel when finished

5. Predicting Ionic Conductivity for an Unseen Test Set

In this section, we will use the AutoQSAR panel to predict ionic conductivity on an unseen test data set of ionic liquids that were not used to train and evaluate AutoQSAR models.

Figure 5-1. Selecting the test set file.

Let’s import the test data set:

  1. Go to File > Import Structures
  2. Navigate to where you downloaded the tutorial files (presumably your working directory) and choose fixed_temp_298_K_test.csv. Click Open

Figure 5-2. Importing in the test set.

  1. Change the ENTRY TITLE Column to Index
  2. Click OK

Figure 5-3. Renaming the group.

 

A test set of 40 ionic liquid structures and ionic conductivity values have now been imported into the project. These structures were not used in the training and evaluation of the AutoQSAR models.

  1. Rename the group name to test_set

Figure 5-4. Choosing the QSAR task.

  1. In the entry list, select(1) the atoms are chosen in the Workspace. These atoms are referred to as "the selection" or "the atom selection". Workspace operations are performed on the selected atoms. (2) The entry is chosen in the Entry List (and Project Table) and the row for the entry is highlighted. Project operations are performed on all selected entries the test_set group
  2. Go to Tasks > Materials > Informatics > AutoQSAR
    • The AutoQSAR panel opens
    • In addition to using the AutoQSAR panel to build and evaluate our models, we can also use this panel to make predictions
  3. For Choose task select View model and make prediction
  4. Ensure that the train_autoqsar.qzip file is still selected

Figure 5-5. Running the job.

  1. Change the AutoQSAR Prediction to Ionic_Conductivity
    • This will be name of the predicted values property

Note: Model to test is set to All models (consensus prediction). Consensus prediction averages the results of the retained models, which can often increase the accuracy of the predictions.

 

  1. Change the Job name to test_autoqsar
  2. Adjust the job settings () as needed
    • This job requires a CPU host and can be completed in about 30 minutes
  3. If you would like to run the job yourself, click Run. Otherwise, import the pre-generated Section_05 > test_autoqsar > test_autoqsar-out.mae.gz file from the provided tutorial files via File > Import Structures
  4. Close the AutoQSAR panel

 

Figure 5-6. Viewing the output of the prediction task.

When the job is complete or after importing, a new entry group is added to the entry lista simplified view of the Project Table that allows you to perform basic operations such as selection and inclusion titled test_autoqsar-out (40) containing the same 40 entries, but now with predicted ionic conductivity properties.

Figure 5-7. The Project Table.

The resulting data can be analyzed in the Project Tabledisplays the contents of a project and is also an interface for performing operations on selected entries, viewing properties, and organizing structures and data

  1. Open the Project Tabledisplays the contents of a project and is also an interface for performing operations on selected entries, viewing properties, and organizing structures and data ()
  2. Use the Property Tree () to include the Ionic Conductivity score (Check the properties of interest under All > Canvas > Secondary > Ionic Conductivity, S/m Liquid)

 

The Project Table should show “Ionic Conductivity” and predictions by AutoQSAR (“Pred Ionic Conductivity”). Approximate prediction uncertainties and domain scores from AutoQSAR are also shown as “Pred Ionic Conductivity SD” and “Predicted Ionic Conductivity Domain Score”, respectively (high magnitudes of domain scores tell you whether a molecule is extremely distinct from the original training set).

 

Figure 5-8. Creating a scatter plot.

  1. Click the Manage Plots button ()
  2. Click Create

Figure 5-9. Plotting the ionic conductivity.

  1. In the new scatter plot, change the following parameters:
    • X-axis: Ionic Conductivity, S/m Liquid
      1. This is the experimental ionic conductivity
    • Y-axis: Pred Ionic Conductivity
      1. This is the predicted ionic conductivity
    • Check Best fit line

The best fit line between predicted and actual values shows a reasonable R2 of 0.84 (an ideal model would have an R2 of 1.00). The results suggest that the ML model derived could be used to predict ionic conductivity values for ionic liquids. Furthermore, this workflow highlights the computational efficiency achieved when using ML approaches as compared to other computational (e.g. ab initio calculations) or experimental approaches. While this tutorial uses a relatively small dataset, one could envision a larger training set would further improve prediction accuracy.

6. Conclusion and References

In this tutorial, we learned how to use the AutoQSAR panel to build machine learning models to predict the ionic conductivity for ionic liquids. These machine learning models enable fast screening of ionic liquids for high ionic conductivities, which could help build safer and more efficient batteries.

For further learning:

For introductory content, focused on navigating the Schrödinger Materials Science interface, an Introduction to Maestro for Materials Science tutorial is available. Please visit the materials science training website for access to 70+ tutorials. For scientific inquiries or technical troubleshooting, submit a ticket to our Technical Support Scientists at help@schrodinger.com.

For self-paced, asynchronous, online courses in Materials Science modeling, including access to Schrödinger software, please visit the Schrödinger Online Learning portal on our website.

For some related practice, proceed to explore other relevant tutorials:

For further reading:
  • The dataset was extracted from the supplementary information of:
    Conductivity prediction model for ionic liquids using machine learning.     DOI: 10.1063/5.0089568
  • The original dataset is from: NIST IL Thermo Database
  • Developing machine learning models for ionic conductivity of imidazolium-based ionic liquids. DOI: 10.1016/j.fluid.2021.113208
  • A generalized machine learning model for predicting ionic conductivity of ionic liquids. Molecular Systems Design and Engineering. DOI: 10.1039/D2ME00046F
  • DeepAutoQSAR: Scalable, Intuitive, Deep-learning QSAR models for Big Data Applications (Schrödinger white paper)
  • DeepAutoQSAR Hardware Benchmark (Schrödinger white paper)
  • See the help documentation for more information on the AutoQSAR panel  

7. Glossary of Terms

Entry List - a simplified view of the Project Table that allows you to perform basic operations such as selection and inclusion

Included - the entry is represented in the Workspace, the circle in the In column is blue

Project Table - displays the contents of a project and is also an interface for performing operations on selected entries, viewing properties, and organizing structures and data

Recent actions - This is a list of your recent actions, which you can use to reopen a panel, displayed below the Browse row. (Right-click to delete.)

Scratch Project - a temporary project in which work is not saved, closing a scratch project removes all current work and begins a new scratch project

Selected - (1) the atoms are chosen in the Workspace. These atoms are referred to as "the selection" or "the atom selection". Workspace operations are performed on the selected atoms. (2) The entry is chosen in the Entry List (and Project Table) and the row for the entry is highlighted. Project operations are performed on all selected entries

Working Directory - the location where files are saved

Workspace - the 3D display area in the center of the main window, where molecular structures are displayed