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

Jaroslaw Polanski

doi:10.3390/ijms23052797

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.

Highlights

The availability of computers has brought novel prospects in drug design
Because cheminformatics currently attempts to organize all of the research that connects chemistry and computer science, we often forget that drug design was its first task
The development of efficient Deep Learning (DL) methods can be observed in deep face and deep blue recently

Summary

Introduction

The availability of computers has brought novel prospects in drug design. The tautological term rational drug design (irrational design would be contrary to logic), coined for computer technologies, illustrates the high expectations in this area. Neural networks (NN) were an early tool that cheminformatics tested for converting data into drugs. In drug design, unsupervised architectures can be surprisingly broadly observed, indicating the efficiency of the method and the fact that we need to process sizeable molecular data when measured properties are not available. This publication reviews recent applications of the DL algorithms for drug design, comparing them to the early unsupervised neural networks. In unsupervised learning applications for mapping molecular representations, we can still recognize the early neural network protoplasts. Machine learning is a method of data processing by various in silico algorithms. When discussing the perspectives of automated drug design and discovery, Schneider signifies the importance of DL methods by indicating their pattern recognition capabilities, especially when patterns escape the medicinal chemistry rationale [2].

From Chemical Compounds to Drugs and Materials

Self-Organizing Mapping of Molecular Representations

Deep Learning for Processing Molecular Data in Drug Design

Conclusions