The question of molecular similarity is core in cheminformatics and is usually assessed via a pairwise comparison based on vectors of properties or molecular fingerprints. We recently exploited variational autoencoders to embed 6M molecules in a chemical space, such that their (Euclidean) distance within the latent space so formed could be assessed within the framework of the entire molecular set. However, the standard objective function used did not seek to manipulate the latent space so as to cluster the molecules based on any perceived similarity. Using a set of some 160,000 molecules of biological relevance, we here bring together three modern elements of deep learning to create a novel and disentangled latent space, viz transformers, contrastive learning, and an embedded autoencoder. The effective dimensionality of the latent space was varied such that clear separation of individual types of molecules could be observed within individual dimensions of the latent space. The capacity of the network was such that many dimensions were not populated at all. As before, we assessed the utility of the representation by comparing clozapine with its near neighbors, and we also did the same for various antibiotics related to flucloxacillin. Transformers, especially when as here coupled with contrastive learning, effectively provide one-shot learning and lead to a successful and disentangled representation of molecular latent spaces that at once uses the entire training set in their construction while allowing “similar” molecules to cluster together in an effective and interpretable way.

This study presents a novel approach to predict a complete source to destination trajectory of a vehicle using a partial trajectory query. The proposed architecture is scalable to extremely large-scale data with respect to the dense road network. A deep learning model Long Short Term Memory (LSTM) has been used for analyzing the temporal data and predicting the complete trajectory. To handle a large amount of data, clustering of similar trajectory data is used that helps in reducing the search space. The clusters based on geographical locations and temporal values are used for training different LSTM models. The proposed approach is compared with the other published work on the parameters as Average distance error and one step prediction accuracy The one-step prediction accuracy is as good as 81% and Distance error are .33 Km. Our proposed approach termed Clustered LSTM is outperforming in both the parameters when compared with other reported results. The proposed solution is a clustering-based predictive model that effectively contributes to accurately handle the large scale data. The outcome of this study leads to improvise the navigation systems, route prediction, traffic management, and location-based recommendation systems.