Customer segmentation logo

Drug design using machine learning

By Pablo Martin, Artelnics.

A technique widely used by pharmaceutical companies in the drug discovery process is called structure-activity relationship (SAR) analysis. A successful solution to this problem has the potential to provide significant economic benefit via increased process efficiency. Predictive analytics techniques such as deep learning have demonstrated to outperform its competitors in predicting the activity of a molecule from the values of its physicochemical descriptors.

Artificial neural networks utilize deep and specialized architectures to learn useful features from raw data, which can be specially useful in nanotechnology. The recent success of machine learning provides an opportunity to develop tools for automatically extracting task-specific representations of chemical structures. Due to this success, they have recently applied to biomedical field.


Study the surface properties, molecular volumes or molecular interactions.
Relate the orientation of the molecule to their characteristics.
Develop a predictive model capable of predicting the behavior of a molecule in accordance to their design.
Get better results and design better medicines while reducing the costs.

The general protocol for constructing SAR models for drug discovery has been systematized and consists of several modular steps involving the chemoinformatics and machine learning techniques. The first step is deriving the chemical features and properties from chemical structures. Second, a feature selection step is performed to identify the most relevant properties and reduce the dimensionality of the feature vector. Finally, in the learning phase, a neural network model is applied to discover a function that can achieve an optimal mapping.

Drugs picture

Neural Designer is a machine learning software that implements neural networks for the analysis of data sets. It can find complex relations between different variables and manage large amount of information. The information that it receives is arranged in a data set which will contain the information about the structure of the molecules.

Neural Designer also provides you with the statistical methods to study the characteristics of the molecules, reduce dimensionality of the data set, design the predictive model and test it against real cases to assess its quality.

The next image shows a summary of the process that would be followed to obtain a predictive model for this case. Firstly, the data about particles are arranged in a database. This database is analyzed by Neural Designer to find patterns. As a result, we obtain a predictive model capable of designing new structures.

Activity diagram

Related solutions:

> Medical diagnosis.
> Microarray analysis.
> Medical prognosis.