Artificial Intelligence for Drug Discovery market is expected to be US$ 8,419 Million by 20261. This traction can be attributed to the growing conviction in the potential of Artificial Intelligence (AI) in reimagining Drug Discovery. Traditional R&D efforts often take 11-15 years with the involved cost being above $2.6 billion2. In addition, nine out of ten drug candidates fail between phase I trials and regulatory approval, making this process ineffective and expensive.
Leading biopharmaceutical companies’ belief in AI is due to growing awareness related to AI in the pharmaceutical sector and rising investment in drug development. AI works with the latest advanced biology and chemistry processes to develop state-of-the-art algorithms; it has the potential to move drug screening from bench to virtual lab without the need of extensive experimental input and manpower.
Most sizeable biopharmaceutical companies have started internal programs or are collaborating to develop AI platforms for increasing the research power of immuno-oncology drugs, metabolic-disease therapies, cancer treatments, and many other therapeutic targets.
AI is changing drug discovery
Applications of AI in drug discovery are extensive and can be classified into following areas:
1) Target selection and validation
Target Identification deals with identification of function of possible molecular target (genes/proteins of a small molecule) and its role in a disease, which is aimed at finding the efficacy target of a drug. This requires evaluation of functional genomics, structural genomics, proteomics, and cell based assays (in-vitro), animal research (in-vivo) assays.
AI is being used to analyse Drug Information Bank (contains drug candidates, gene expressions, protein-protein interactions and clinical data records) from a public library for predicting therapeutic potential. For example, applying feature engineering by deep autoencoder, relief algorithm and binary classification by Xgboost algorithm on ‘genome-wide protein interaction network, drugs and their targets information’ to produce scores for potential targets to enable target prioritization.
For finding of drug target site, discrete chemicals can be encoded into continuous latent vector space, which permits gradient-based optimization in molecular space, and allows predictions using graph convolutional network based on binding affinity and others properties.
In the effort to understand the detailed spatial 3D structure of proteins and molecular complexes, AI platform also rely on training computer vision and machine learning models on cryo-EM microscope data (2D structure).
Selection of drug candidates requires series of desired properties especially pharmacokinetics, pharmacological and toxicity profile. AI drug design algorithm Reinforcement Learning can be successfully deployed for a ‘Simplified molecular input line-entry system’string, which is a specification in the form of line notation for describing Chemical species using short ASCII strings. It includes potential energy measurements, molecular graphs with varying weights for atoms or bonds, coulomb matrices, molecular fragments or bonds, atomic coordinates in 3D etc.
2) Compound screening and lead optimization
Compound screening and lead optimization step involves Hits followed by Leads identification, where selection of drug candidates are done through combinatorial chemistry, high throughput screening, and virtual screening.
AI-based Virtual Screening is the compound database that is made through pulling mass quantities from publicly available chemogenomics libraries, which includes tens of millions of compounds annotated with information about their structure. It uses Naïve Bayesian Classifiers, k-Nearest Neighbors, Support Vector Machines, Random Forests and Artificial Neural Networks algorithms to enable Medicinal Chemists to efficiently find potential lead molecules among millions of compounds, thus speeding up initial stages of drug development.
Planning Chemical Synthesis with AI Retrosynthesis Pathway Prediction are ambitious plans to exploit AI to automate chemical syntheses with minimal manual operation. Synthesis robots combined with AI can be used here. AI platform named 3N-MCTS, which combines three different deep neural networks with Monte Carlo Tree Search for computer-aided organic compound synthesis, can select only well-known reactions by filtering out most promising building blocks, for the synthesis of target compounds.
For cell target classification, the AI model has to be trained to quickly and automatically recognize the different features of cell types. Principal component analysis is employed to reduce the dimensions of the extracted features. AI-based methodologies such as the least-square support vector machine can be trained to classify various cell types.
To accurately separate different cell types in the sample during Cell sorting step, Intelligent Image-Activated Cell Sorting devices are helpful, which measures optical, electrical, and mechanical cell properties through AI-based convoluted deep neural network algorithms.
3) Preclinical studies
Preclinical Studies or non-clinical studies are laboratory tests for new drug substance in in-vitro and in-vivo for establishment of its efficacy and safety profile.
For establishing ‘molecular mechanism of action’, unsupervised approach of clustering based machine learning tools evaluates RNA sequencing technologies, this in-turn shortens the time to capture relevant significant quantities of biological information. This also leads to uncover dozens of previously unexplored links between diverse stimuli and cytokines they effect.
Pharmacokinetic/Pharmacodynamic ML Modelling approach, works on in-vitro studies and preclinical PK studies to predict ‘dose concentration (exposure) response relationship’. ML models are also being used in Drug-Dose response for efficacy predictions which can be used for effective multidrug combinations using small number of experiments.
In evaluation of toxicology profile of a compound, the most expensive parameter and time consuming task, Deeptox Algorithm is already evaluated for more than 10000 environmental chemicals and drugs for 12+ different toxic effects in explicitly designed assays 3. Hence, it can add great value to drug development by accurately predicting the toxicity of compounds.
Deep Learning Algorithms are used in ‘In-Silico’ methods for predicting pharmacological properties by using transcriptomic data which includes various biological systems and conditions.
4) Clinical trials
Development of AI tool for clinical trials would be ideal for recognizing diseases of patients, identifying gene targets, predicting the effect of molecule designed, and on and off targets. One such mobile application AI platform, when compared to traditional ‘modified directly observed therapy’ has increased medication adherence by 25% in Phase II clinical trials4.
AI in risk based monitoring, which is a clinical trial monitoring technique that fulfills regulatory requirements but moves away from 100% source data monitoring, can significantly improve the conduct of clinical trials in all phases. In Phase II and III of clinical trials, AI can be used to identify and predict human-relevant biomarkers of disease to select and recruit specific patient population, which would lead to increase in success rate of clinical trials.
Technical obstacles and future prospects
The biggest limitation in using AI to predict drug targets remains in translating traditional basic research performed in labs around the world into a language that a computer perceives. However, in many cases, the data being used are not of optimal quality or are not balanced. While data augmentation techniques and improving image quality and variation have been largely studied, these remain a challenge.
Patent of AI solution for drug discovery has to go through an intense process. Security of medical/drug discovery data is also a big challenge and proper security measures are very essential.
In the present scenario, it is a challenge to comprehensively test a new lead compound in combination with all available drug molecules. It takes thousands of studies to analyze known side effects and unknown interactions. However, once available, such an AI algorithm approach, would prove invaluable in further hastening drug development efforts.
AI will revolutionize how drugs are discovered and will reinvent the pharmaceutical industry.
- Mayr, A. et al. (2016) DeepTox:toxicity prediction using deep learning. Front.Environ.Sci. 3, 80
- Bain, E.E. et al. (2017) Use of a novel artificial intelligence platform on mobile devices to assess dosing compliance in a Phase 2 clinical trial in subjects with schizophrenia. JMIR Mhealth Uhealth 5, e18 Aicure Clinical trials