Indrani Raman Mahadik
Second Year M. Pharm(Chemistry)
How machine learning can be used to find new therapeutic uses for already-approved medications, potentially cutting down on the time and expense of drug development.
The process of discovering new therapeutic uses for existing drugs that have already been approved for a different indication is known as drug repurposing. Because of its potential to reduce the time and costs associated with drug development, this approach has gained popularity in recent years. In this blog post, we will look at how machine learning can be used to discover new therapeutic applications for existing drugs.
One approach to drug repurposing using machine learning is to use computational models to predict the efficacy of existing drugs in treating various diseases. A machine learning algorithm, for example, can be trained on a dataset of gene expression data from patients with a specific disease, as well as information about the drugs used to treat that disease. Based on known mechanisms of action and observed gene expression patterns, the algorithm can then use this data to identify drugs that are likely to be effective in treating the disease.
Another approach to drug repurposing with machine learning is to identify new drug-disease associations using network-based methods. Drugs and diseases are represented as nodes in a network in this approach, with edges representing known interactions between them. Machine learning algorithms can then be used to identify novel drug-disease associations based on the network structure and node properties.
By analysing large amounts of data and identifying potential new therapeutic uses for existing drugs, machine learning, a subset of artificial intelligence, can play a critical role in drug repurposing. Here are some examples of how machine learning can be applied to drug repurposing:
Identifying subtypes of diseases: Based on clinical and molecular characteristics, subtypes of diseases can be identified with machine learning. This can assist in locating currently available medications that might be useful for treating a specific subtype of a disease.
Bringing together data from various sources: Integrating data from a variety of sources, including clinical trial data, genomics data, and electronic health records, is one way that machine learning can find new therapeutic applications for existing drugs.
Putting drug candidates first: Prioritize existing drugs based on their potential for repurposing with machine learning. This can assist researchers in concentrating on drugs that are most likely to treat a specific disease.
New clinical trial design: New clinical trials for repurposed drugs can be designed with the help of machine learning. These algorithms can assist in determining the most promising repurposed drugs for clinical trials by analyzing data on patient characteristics and disease subtypes.
Recent instances of AI-enabled drug repurposing successes
A number of recent studies have demonstrated machine learning’s potential for drug repurposing. A machine learning algorithm was used in a 2020 Nature Communications study to predict the therapeutic potential of 1,309 FDA-approved drugs for 55 different cancers. The antipsychotic drug thioridazine, which was found to be effective against acute myeloid leukemia, was one of several drugs that the algorithm identified that showed promise for treating particular types of cancer.
Another example is a study that used machine learning to find potential drugs that could treat COVID-19 and was published in Nature Biotechnology in 2021. A neural network algorithm was used by the researchers to analyze the molecular structures of existing drugs and predict their capacity to bind to the SARS-CoV-2 spike protein, a vital virus component. The algorithm found several drugs that could treat COVID-19, including the antidepressant fluvoxamine and the antiparasitic niclosamide.
To summarise, machine learning has the potential to transform drug repurposing by identifying novel therapeutic uses for existing pharmaceuticals in a more efficient and cost-effective manner. However, the quality of the data used to train the algorithms must be carefully monitored, and the findings must be interpreted with sensitivity to ensure that they are biologically meaningful.