ChemXploreML

Appearance

Molecular Embeddings ​

The "Embed Molecule" feature in ChemXploreML is used to convert molecular structures from SMILES strings into numerical vectors, a process known as embedding. These embeddings are essential for machine learning tasks, as they provide a quantitative representation of the molecules.

Overview ​

Vectorize Molecules Interface

This module allows you to generate embeddings using several state-of-the-art models. You can either process an entire dataset ("File Mode") or test a single molecule ("Test Mode").

Supported Embedding Models ​

ChemXploreML supports the following pre-trained models for generating molecular embeddings:

  • VICGAE: A graph-based model for learning molecular representations.
  • mol2vec: An unsupervised machine learning approach to learn vector representations of molecular substructures.
  • ChemBERTa-zinc-base-v1: A transformer-based model pre-trained on a large dataset of molecules.
  • MoLFormer-XL-both-10pct: Another powerful transformer-based model for molecular embeddings.

For each of these models, ChemXploreML can automatically download the necessary pre-trained weights if they are not already available. You also have the option to use your own pre-trained model files.

Workflow ​

1. Test Mode ​

If you want to quickly see the embedding for a single molecule:

  1. Enable the Test mode checkbox.
  2. Enter a SMILES string in the input field.
  3. Select an embedder from the dropdown menu.
  4. Click the "Run" button.
  5. The resulting numerical vector will be displayed in the text area.

2. File Mode ​

To generate embeddings for an entire dataset:

  1. Make sure Test mode is disabled.
  2. Ensure you have a file loaded from the "Load File" tab. The selected "Column X" (containing SMILES) will be used.
  3. Select an embedder from the dropdown menu.
  4. Click the "Run" button to start the process.
  5. The application will process the entire file and save the resulting embeddings as a .npy file.

3. Results ​

After the embedding process is complete, a summary of the results will be displayed, including:

  • The dimensions of the generated embedding matrix.
  • The number of any invalid SMILES that could not be processed.
  • A link to a file containing the invalid SMILES for your review.

Next Steps ​

Once you have generated the molecular embeddings, you are ready to use them for: