Issue link: https://www.e-digitaleditions.com/i/1509086
38 Pharmaceutical Technology ® Trends in Formulation 2023 eBook PharmTech.com CPHI BarCelona Coverage might not be able to easily deal with these types of data. Quantitative structure-activity relationship (QSAR)- based computational models can quickly predict large numbers of compounds or simple physicochemical parameters but have not matured to the point where they can accurately predict efficacy and adverse effects of compounds. QSAR has been integrated with deep learning (DL) modeling to address its limitations and the combination has successfully screened compounds, addressing efficacy and toxicity considerations, with promising results. Furthermore, artificial neural net- works (ANN) coupled with QSAR have been found to be effective in pushing the capabilities of in-silico pre- dictive power and accuracy. The effectiveness of drug screening using AI will improve as we build and refine the training data set for evaluation. Drug design. One of the primary challenges in devel- oping a drug molecule is identifying the correct target for treatment. Numerous proteins are involved in the development of a disease target, and in some cases, the proteins are overexpressed. To selectively target a disease, one must predict the structure of the target protein for designing the drug molecule. AI can assist in structure-based drug discovery by predicting the 3D protein structure. Understanding the 3D structure allows drug sponsors to not only predict the effect of a compound on the target but also will help identify safety considerations before their synthesis or produc- tion. AI can assist in predicting drug–protein interac- tions as part of determining efficacy and effectiveness. This can lower program risk for drug sponsors looking to repurpose existing drugs and reduce the likelihood of a drug molecule interacting with multiple protein receptors producing off-target adverse effects. AI pro- vides two distinct advantages to drug sponsors from a drug design perspective by eliminating drug designs that have a higher risk of safety or efficacy issues early in the development program and isolating designs that have a higher probability of success. Getting to a "no-go" decision quickly is just as valuable as putting the time and money into a drug with a high probability of success. A computational de novo drug design approach can leverage AI. The traditional method of de novo drug design is being replaced by evolving DL methods. With computer-aided synthesis, it is possible to suggest mil- lions of structures that can be synthesized as well as predict different synthesis routes for them. Multiple AI platforms have demonstrated a supe- rior effect when compared to the trial-and-error ap- proach traditionally applied to drug design, reducing both time required and probability of an unexpected adverse response. AI in product development. AI is being used to ad- dress product formulation issues and requirements, in- cluding stability issues, dissolution, porosity, and so on, with the help of quantitative structure–property rela- tionship (QSPR) analysis. QSPR is an approach intended to find correlations between material properties and predefined structural descriptors, through regression or machine-learning approaches. Coupled with decision support tools, the development exercise is able to select the type, nature, and quantity of the excipients depend- ing on the physicochemical attributes of the drug. Com- bining AI with established characterization techniques such as computational flow dynamics, discrete element analysis, and finite element analysis has the potential to rapidly characterize and optimize the formulation and product development exercise. This analysis can be framed to optimize patient cen- tricity as well. Considerations such as the route of ad- Clinical Trials AI is being used to identify target patient cohorts quicker and simplify the engagement process for patients.. Drug Screening Finding a novel molecule with the right balance of on-target affinity and desired physicochemical properties considering key factors such as toxicity and bioactivity . Process/Manufacturing AI is being used to characterize unit operations and establish the control strategy for manufacturing. Product Development AI is being used to address product formulation issues and requirements including stability issues, dissolution, porosity. QA and QC Artificial intelligence (AI) is revolutionizing pharmaceutical Quality Control by enhancing the accuracy and speed of inspections. Drug Design Computer-aided synthesis makes it possible to suggest millions of structures AI Applications in Pharma FIGURE 1. Artificial intelligence (AI) applications in pharma and biotech. FIGURE COURTESY OF THE AUTHOR.