Challenges to the adoption of Artificial Intelligence (AI) in Pharma

drug manufacturing

Historically, the discovery of new drugs takes a long time to extract ingredients from natural products, test compounds against samples of diseased cells, and find an appropriate candidate worth exploring in more detail.

The drug discovery process usually takes five to six years from the start to the end of preclinical testing. Of 10,000 small molecules initially screened, 10 are selected for clinical trials. This progress is generally long, expensive, labor-intensive, and often unsuccessful.

Thankfully, the recent advancements in technology significantly speed up every step in drug development. Today, pharmaceutical companies can benefit substantially from various machine learning (ML) techniques in predicting chemical, biological, and physical characteristics of compounds in drug discovery.

These tools allow computer systems to perform tasks requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages. They are far quicker compared to traditional human analysis and laboratory experiments.

With humongous data available in the drug R&D process, artificial intelligence (AI) can help analyze the data and present results that would help in decision-making, saving human effort, time, and money, associated with the manual investigation of each compound.

In general, drug companies use supervised and unsupervised learning methods in machine learning, such as Random Forest (RF), Naive Bayesian (NB), and support vector machine (SVM), to predict drug-protein interactions, discover drug efficacy, ensure safety biomarkers, and optimize the bioactivity of molecules.

Other commonly used ML techniques and models include regression, clustering, regularization, neural networks (NNs), decision trees, dimensionality reduction, ensemble methods, rule-based methods, and instance-based methods.

The immense benefits of using AI in drug discovery are as follows:

  • Quick identification and optimization of potential new drugs.
  • Reduce timelines for the drug discovery process and improve the agility of the research process.
  • Increase the accuracy of predictions on the efficacy and safety of drugs
  • Improve the opportunity to diversify drug pipelines, open up new research lines, and more competitive R&D strategies.

Despite the promise of artificial intelligence and machine learning to transform the pharmaceutical industry, putting these technologies into practice comes with its own set of challenges. Some of the challenges that pharma companies face while trying to adopt AI are:

  • The unfamiliarity of the technology: For many pharma companies, AI still seems like a “black box” owing to its newness and esoteric nature.
  • Lack of proper IT infrastructure: Most IT applications and infrastructure currently in use are not developed or designed with AI in mind. Even worse, pharma companies have to spend lots of money to upgrade their IT system.
  • Breaking down data silos and streamlining electronic records: Much of the data is in a free text format, and the data management is messy and unorganized across the heterogeneous databases. This means that pharma companies have to go above and beyond to collate and put this data into a standard form that can be analyzed.
  • Low accuracy of the training data: Even though algorithms have a higher threshold for minimizing errors, there are still some categorical errors from training sets.
  • Overfitting or underfitting: With algorithm prediction, there is a concern with overfitting or underfitting. Overfitting means when a model consists of lower quality information/technique but generates higher quality performance. Underfitting models fail to recognize the underlying trend in the datasets and generalize the new data. Both result in inaccurate results.
  • Data quality, governance, security, and interoperability: Issues around data will always be at the heart of successfully promoting AI solutions. Healthcare is the least digitized sector, which needs to take a systematic approach to develop common data standards and processes to maximize the value of existing data. Healthcare providers and AI companies need to put in place robust data governance, ensure interoperability and standards for data formats, enhance data security and bring clarity to consent over data sharing.
  • The need for transparent algorithms to meet drug development regulations: Transparency in health care is quite a task given the complexity of the processes involving artificial intelligence.
  • Hesitant to change: Pharma companies are known to be traditional and resistant to change.