Drug Research and Development
The costs of bringing a new drug to market is a long, costly, and complex process that can often take more than 10 years from identification of molecular targets until a drug product is approved and marketed. According to a new study led by Dr Olivier Wouters, Assistant Professor of Health Policy at the London School of Economics, the median cost of bringing a new drug to market is $985 million, to an average cost of $1.3 billion. Thus, any failure during this process has a large financial impact, and in fact most drug candidates fail sometime during development and never make it onto the market.
In addition, there are the complicated regulatory hurdles to overcome and the ever continuing difficulties in discovering drug molecules that are substantially better than what has been discovered before. This makes the drug innovation process both challenging and inefficient with a high price tag on any new drug products that eventually make it onto the market. However, since the 1990s there has been a substantial increase in the amount of data available assessing drug compound activity and biomedical data.
This is due to the increasing automation and the introduction of new experimental techniques including hidden Markov model based text, to speech synthesis and parallel synthesis. But, mining of the large-scale chemistry data is needed to efficiently classify potential drug compounds and machine learning techniques have shown great potential in this area. More recently, Deep Learning (DL) has begun to be implemented due to the increased amount of data, and the continuous improvements in computing power.
Machine learning can be utilised to streamline the process at various stages in the drug discovery process. This includes drug compound property and activity prediction, de novo design of drug compounds, drug–receptor interactions, and drug reaction prediction. For example, the properties and activity on a drug molecule are important to know in order to assess its behavior in the human body. Machine learning-based techniques have been used to assess the biological activity, absorption, distribution, metabolism, and excretion characteristics, and physicochemical properties of drug molecules.
In recent years, a number of libraries storing information on millions of molecules for various disease targets have also become available. These libraries are machine-readable and are used to build machine learning models for drug discovery. For instance, convolutional neural networks have been used to generate molecular fingerprints from a large set of molecular graphs with information about each atom in the molecule. Neural fingerprints are then used to predict new characteristics based on a given molecule.
To conclude, there is ongoing work to reduce the amount of data required as training sets for DL, so it can learn with only small amounts of available information. This is similar to the learning process that takes place in the human brain, and would be beneficial in applications where data collection is resource intensive and large datasets are not readily available, as is often the case with medicinal chemistry and novel drug targets.
