Fig. 6
From: Positional embeddings and zero-shot learning using BERT for molecular-property prediction
Analysis of the similarity between the pretraining and downstream (i.e., Clintox) SMILES dataset samples (i.e., 100 randomly selected samples) based on the molecular structures using BERT with “Relative_key” PE. a Tanimoto similarity heatmap, b Embedding similarity heatmap, and c PCA Visualization of molecular embeddings in latent space