Unsupervised Learning in Drug Design from Self-Organization to Deep Chemistry

Abstract

The availability of computers has brought novel prospects in drug design. Neural networks (NN) were an early tool that cheminformatics tested for converting data into drugs. However, the initial interest faded for almost two decades. The recent success of Deep Learning (DL) has inspired a renaissance of neural networks for their potential application in deep chemistry. DL targets direct data analysis without any human intervention. Although back-propagation NN is the main algorithm in the DL that is currently being used, unsupervised learning can be even more efficient. We review self-organizing maps (SOM) in mapping molecular representations from the 1990s to the current deep chemistry. We discovered the enormous efficiency of SOM not only for features that could be expected by humans, but also for those that are not trivial to human chemists. We reviewed the DL projects in the current literature, especially unsupervised architectures. DL appears to be efficient in pattern recognition (Deep Face) or chess (Deep Blue). However, an efficient deep chemistry is still a matter for the future. This is because the availability of measured property data in chemistry is still limited.

Retrosynthesis from transforms to predictive sustainable chemistry and nanotechnology: a brief tutorial review

“…We gain more and more experience in backpropagation routines; however, unsupervised architectures could be even more efficient. 53 Beyond synthesis design, drug discovery is another area of molecular design widely explored by machine learning. 40,41 The reader can find an extensive up-to-date review of machine learning for molecular and materials sciences in ref.…”

“…Early neural network applications can be illustrative examples for a better understanding of these methods, 54,55 in particular, a comparison of supervised vs. unsupervised architectures in deep chemistry is presented in ref. 53.…”

“…We gain more and more experience in back-propagation routines; however, unsupervised architectures could be even more efficient. 53…”

From intuition to AI: evolution of small molecule representations in drug discovery

“…The latent space occupied by these learned embeddings is highly organized. Indeed, impressive clustering of similar molecules at close points in latent space is observed as a result of unsupervised training [ 95 ]. While autoencoders and transformers can encode molecules into latent space through their encoder modules, they can also decode latent space vectors back into small molecules [ 96 ].…”

DrugGPT: A GPT-based Strategy for Designing Potential Ligands Targeting Specific Proteins

“…While traditional computational drug design strategies such as molecular docking (Crampon et al, 2022), Quantitative Structure-Activity Relationship (QSAR) (Muratov et al, 2020), pharmacophore modeling (Pautasso et al, 2014), unsupervised learning (Polanski, 2022), deep learning (Wang et al, 2022), and Generative Adversarial Networks (GANs) (Tong et al, 2021)have somewhat mitigated this issue, the demand for more effective exploration of this immense chemical space necessitates the development of innovative approaches and strategies (Öztürk et al, 2020).…”