Artificial Intelligence and Pharmacogenomics

Sophia Z. Humphreys, PharmD, MHA, BCBBS, CPGx is Executive Director, Pharmacy at Providence Health, and Clinical Assistant Professor at University of Washington. With 20+ years’ expertise in clinical practice, formulary strategy, utilization management, Pharmacoeconomics, and biosimilar adoption, she is a recognized speaker and published author.

In an exclusive interview with CIOReview, she shared invaluable insights on how AI is transforming pharmaceutical care by enabling more efficient, accurate, and personalized treatment decisions.

From Trial-and-Error to Tailor-Made: AI’s Role in Drug Therapy

Artificial intelligence (AI) is rapidly reshaping the field of pharmaceutical care, introducing transformative changes across various facets of drug development, medication management, and patient engagement. AI serves as a powerful tool that augments scientific capabilities, allowing healthcare professionals to provide more efficient, accurate, and personalized care. AI tools can assist pharmacists to manage medication inventory, predict drug demand, identify potential drug-drug interactions, and optimize drug dosages. AI-assisted decision making can help design treatment regimens based on individual patient’s genetic makeup, lifestyle, and medical history, leading to improved outcomes and reduced adverse events.

AI has also revolutionized pharmacogenomics. By analyzing complex genetic data and integrating it with clinical information, AI can help create personalized treatment plans, optimize drug selections, and improve patient outcomes. This includes predicting how a patient will respond to certain medications based on genetic makeup, leading to more effective and safer drug therapies. For example, genetic polymorphisms in enzymes responsible for medication metabolism may lead to notable interpersonal variations in response to a medication. This may include varied toxicity levels and adverse events to the same dosage (1). Some mutations may reduce the activities of the primary enzymes which metabolize a certain drug or a subclass of drugs, thus causing increased toxicity in some patients (i.e., the poor metabolizer). In other situations, when a patient has genes coded for the enzymes that metabolize the same drugs at a much faster rate, we identify them as rapid or super-rapid metabolizers. These patients would require a much higher dosage, compared to normal patients, to achieve optimal therapeutic effects.

The most common enzymes known to influence drug metabolism are cytochrome P450 2D6, 3A4, 2C9, 2C19 and 3A5. Mutations also have been found in genes involved in drug conjugation (like N-acetyltransferase 2 or NAT2) (2).

CYP 2D6 is heavily involved in the metabolism of Tamoxifen, antidepressants, codeine, and some pain medications. There are many variants which may lead to reduced or increased 2D6 activity. This can increase toxicity for poor metabolizers and reduce drug levels in rapid or ultra-rapid metabolizers who are taking the same medications. As a result, certain patients may need much more or less than the usual medication dosage to achieve a therapeutic serum drug level (2).

The Genetic Code Meets Machine Intelligence in Healthcare

AI tools can assist healthcare providers by assessing the existing data in the Electronic Health Record (EHR), cross-referencing with published scientific information, and predicting how a patient with a specific type of cytochrome P450 enzyme mutation may need a higher, normal, or lower dosage for a particular medication. This can optimize drug efficacy, minimize toxicity, and predict whether patient will respond well to certain medications.

  ​Artificial intelligence empowers pharmacists to analyze genetics and clinical data, predict drug responses, optimize dosages, and create precise treatments that improve outcomes while reducing risks   

For example, an ultrarapid metabolizer could take the same dosage as a slow metabolizer. The slow metabolizer would display toxicity related side effects, but the ultrarapid metabolizer would not show any clinical benefits. In traditional practice, a pharmacist would need to collect multiple blood tests for drug levels or change dosages to determine if the patient is a suitable candidate for the treatment. However, AI tools, together with accurate pharmacogenomic data, drug information and patient specific genetic information, would be able to predict the best dosage for each medication for each individual patient. This is especially important for cancer treatment, where toxicities may lead to a reduction of dosage, and interruption of treatment. It is not surprising that many healthcare systems are exploring AI tools to enhance pharmacogenomics service for their patients (2).

AI tools can interpret and analyze both population and personalized genetic information to predict patient specific responses to certain medications and recommend the best dosage in each case. For example, a poor metabolizer of a certain medication will need a much lower dose (lower than the normal population) to reach a therapeutic serum level for the medication. This will help the healthcare providers create precise treatment regimens (agents of choice and personalized dosage) according to the patient's specific genetic makeup.

In addition, AI supported decision making tools may combine pharmacogenomic information in the Electronic Health Record (EHR) system with the patients’ clinical data, lifestyle information, family history, and real time labs. This can assist clinicians to make more personalized treatment regimens (3). These tools can also adjust the treatment plans as the patient’s response to the treatment progresses (4).

Oncologists have seen many successes in using AI tools to develop better personalized treatment regimens. The AI-aided systems can identify antineoplastic drugs that have the best clinical outcome and lowest risks based on the patients’ genetic makeup and the types of the cancer the patient has (5). As mentioned before, AI-aided treatment decision making tools can predict a patient’s response to certain medications and the precise dosage the patient would need, based on his/her CYP 450 enzyme genotype and phenotype. This can avoid the trial-and-error type of practice traditionally used for these illnesses (6).

In summary, both genotype and phenotype variations in medication absorption and metabolization are crucial in treatment regimen design. AI-assisted decision making can streamline the analysis and interpretation of the large amount of data within the EHR system and external data sources. They can analyze complex genetic and clinical data and cross-reference external research databases to develop more efficacious and safer treatment regimens. This can reduce the need for trials of less effective and/ or higher risk treatments, thus reducing the total cost of care and utilization of healthcare resources. AI tools can create a winning result for patients, health systems and the financial sustainability of the healthcare industry (7).

References

1. Taherdoost, H., & Ghofrani, A. (2024). Artificial Intelligence and the Evolution of Personalized Medicine in Pharmacogenomics. Intelligent Pharmacy. https://doi. org/10.1016/j.intpha.2024.100011

2. Chiu, M (2024). Using generative AI assistant to interpret pharmacogenetic test results. Baylor College of Medicine. News. March 13, 2024. Accessed on July 29, 2025. Retrieved from: https://www.bcm.edu/news/using-generative-ai-assistant-tointerpret-pharmacogenetic-test-results

3. Chen, YM., Hsiao, TH., Lin, CH. et al. Unlocking precision medicine: clinical applications of integrating health records, genetics, and immunology through artificial intelligence. J Biomed Sci 32, 16 (2025). https://doi.org/10.1186/s12929-024- 01110-w

4. Dhieb, D., & Bastaki, K. (2025). Pharmaco-Multiomics: A New Frontier in Precision Psychiatry. International Journal of Molecular Sciences, 26(3), 1082. https://doi.org/10.3390/ ijms26031082

5. Primorac, D., Bach-Rojecky, L., Vađunec, D., Juginović, A., Žunić, K., Matišić, V., et al. (2020). Pharmacogenomics at the Center of Precision Medicine: Challenges and Perspectives in an Era of Big Data. Pharmacogenomics, 21(2), 141–156. https://doi. org/10.2217/pgs-2019-0111

6. Pardiñas, A. F., Owen, M. J., & Walters, J. T. R. (2021). Pharmacogenomics: A Road Ahead for Precision Medicine in Psychiatry. Neuron, 109(14), 2104–2117. https://doi. org/10.1016/j.neuron.2021.05.030

7. O’Connor, O., McVeigh, T.P. Increasing use of artificial intelligence in genomic medicine for cancer care- the promise and potential pitfalls. BJC Rep 3, 20 (2025). https://doi.org/10.1038/ s44276-025-00135-4