Mol2vec: Unsupervised Machine Learning Approach with Chemical Intuition

Abstract

Inspired by natural language processing techniques we here introduce Mol2vec which is an unsupervised machine learning approach to learn vector representations of molecular substructures. Similarly, to the Word2vec models where vectors of closely related words are in close proximity in the vector space, Mol2vec learns vector representations of molecular substructures that are pointing in similar directions for chemically related substructures. Compounds can finally be encoded as vectors by summing up vectors of the individual substructures and, for instance, feed into supervised machine learning approaches to predict compound properties. The underlying substructure vector embeddings are obtained by training an unsupervised machine learning approach on a so-called corpus of compounds that consists of all available chemical matter. The resulting Mol2vec model is pre-trained once, yields dense vector representations and overcomes drawbacks of common compound feature representations such as sparseness and bit collisions. The prediction capabilities are demonstrated on several compound property and bioactivity data sets and compared with results obtained for Morgan fingerprints as reference compound representation. Mol2vec can be easily combined with ProtVec, which employs the same Word2vec concept on protein sequences, resulting in a proteochemometric approach that is alignment independent and can be thus also easily used for proteins with low sequence similarities.

A natural language processing approach based on embedding deep learning from heterogeneous compounds for quantitative structure–activity relationship modeling

“…In this work, a novel approach for QSAR modeling is proposed, in which we adopt the word embedding approach to predict the activity or the propriety of compounds by using their SMILES strings. Compounds are represented by learning features from a large SMILES corpus via Word2vec (Jaeger, Fulle, & Turk, 2018). Then, we describe each chemical using the average of its interacting ligand vectors that are built by word2vec.…”

Artificial Intelligence in Drug Discovery: A Comprehensive Review of Data-driven and Machine Learning Approaches

“…Neural graph fingerprints [69] mimicked Morgan algorithms [70], bringing atom features in radius, while neural graph fingerprints yielded continuous features of hidden layers to solve sparsity of Morgan fingerprints. Mol2vec [71] generates compound features that can be used to predict bioactivities by applying the Word2vec algorithm on a compound's molecular graph. Simplified molecular-input line-entry system (SMILES) is a well-defined representation of chemical compounds, converting the molecular graph to a sequence of atoms and bonds.…”

“…Each compound is fed into Multi-Layer Perceptron (MLP) and cosine similarity between their latent representation is translated to a Connectivity Map (CMap) score. Therefore, ReSimNet gives a better performance than the conventional machine learning model of ECFP and Mol2Vec [71]. Conversely, the ensemble model of the hierarchical evolutionary chemical binding similarity (ECBS) tree builds more reliable screening results [80].…”

“…Second, most of the studies used engineered molecular descriptors to train the conventional machine learning models, but there are feature-intrinsic biases and a modelinherent low interpretability [72]. To address these problems, a novel data-driven feature generation [71,73] or targetspecific feature learning from raw data offer good solutions [76,136,154,179,184,185]. However, having interpretability in the conventional machine learning and deep learning model is still challenging.…”

A Merged Molecular Representation Learning for Molecular Properties Prediction with a Web-based Service

“…As it is sequential and composed of text, methods inspired by NLP such as word embedding 16 , RNN 17,18 have been proposed. Mol2Vec 19 is a molecular representation inspired by word2vec. It overcomes the drawbacks of fingerprint such as bit collisions.…”