Issue link: https://www.e-digitaleditions.com/i/1173734
Tablets & Capsules October 2019 19 against a predefined set of objectives as it gathers data. Facial recognition is one example of this application. Machine learning is not only being used on the shop floor to optimize performance, applications of machine learning are being applied in drug discovery to improve the success rates of new drug therapies and drug modali- ties as they move through the clinical pipeline. An MIT study published in April 2019 concluded, after analyzing more than 21,000 clinical trials between 2000 and 2015, that only 13.8 percent of drugs suc- cessfully pass clinical trials [2]. This success rate is not sustainable in the face of downward pricing pressures across the globe. One large pharma organization is using machine learning to improve its molecule selection pro- cess. The researchers are building large libraries of digital images of cells treated with different experimental com- pounds and then using machine-learning algorithms to screen potential compounds faster and with a higher rate of success than was previously possible. One potential blockbuster application of AI and machine learning is the treatment of complex diseases that have multiple modes and mechanisms of action, such as autoimmune diseases like multiple sclerosis (MS) or ALS. Currently, research typically targets one gene anomaly or defect. Using AI, researchers may be able to identify multiple genes that influence the disease and devise drug therapies against multiple targets. Another interesting application of AI is happening in clinical treatment. Some cancer treatments are toxic, requiring a complex dosing regimen called dynamic dosing, in which the dose is gradually adjusted to maxi- mal delivery as the patient's treatment progresses. Using AI, doctors can continuously identify each individual curate them, apply security and governance, and make the data accessible for analysis as needed. Such systems now provide pharma with the poten- tial for a single portal and interface to all potential data across the entire business value chain. Most importantly, this doesn't require disassembling any of the solutions that have been put in place. The reluctance to migrate away from legacy systems is one of the biggest organi- zational hurdles faced by cross-functional data manage- ment initiatives. Look for big data initiatives within the industry to shift to these solutions in the next three to five years. Artificial intelligence, real solutions Few areas of innovation have as broad a potential impact as AI. If you think of IoT as connecting devices to gather data, then AI makes decisions based on that data. As such, the applicability of AI is not limited to the shop floor or the manufacturing supply chain. Figure 2 shows the broad applicability of AI across the entire pharma and biotech value chain. Broadly speaking, the term AI applies to any tech- nique that enables computers to mimic human intel- ligence. To fully understand the applicability of AI, it is important to look closer at its two subsets: machine learning and deep learning. Machine learning. Machine learning is the application of targeted statistical techniques that enable machines to improve upon tasks with experience. Machine learning has been used in combination with well-established tech- niques such as "fuzzy logic" to build a set of rules that allows equipment to consistently improve its performance Figure 2 Application of AI across the pharma and biotech value chain Literature Identification Validation Identification Optimization Preclinical Clinical Compliance monitoring Patient monitoring Marketing optimization Target discovery Drug discovery Development Post-approval Scientific data Patient data Healthcare outcomes data Note: Many companies span multiple drug lifecycle stages and data types, therefore relative positioning is indicative Source: L.E.K. research