Can AI drug repurposing really tailor treatments to each patient?
Yes, but only in specific, well-designed studies. A 2026 study on blast-crisis chronic myeloid leukemia (CML) used an AI strategy that analyzed each patient's unique mutational signature from whole-exome sequencing [2]. The result: each of the 12 patients had a distinct therapeutic profile, with no two patients sharing the same recommended drug list. For example, patients with signatures S3/S5 were recommended PARP inhibitors like olaparib, while those with S1 were recommended IDH inhibitors like enasidenib [2]. This demonstrates that AI can move beyond one-size-fits-all approaches to true N-of-1 precision—at least in this cancer type.
What are the main obstacles to using AI across diverse patient groups?
The biggest hurdles are data quality and model interpretability. A 2026 comprehensive review of AI in drug discovery explicitly lists 'patient heterogeneity' as a key translational hurdle, alongside data quality and regulatory adaptation [1]. The review notes that many AI models are trained on datasets that may not represent the full diversity of real-world populations, leading to potential biases [1]. For instance, a 2022 review of structure-based drug repurposing methods points out that traditional AI models often rely on static molecular structures, which may not capture the dynamic biological differences between patients [4]. This means that while AI can work well in controlled studies, its performance may degrade when applied to populations not well-represented in the training data.
Does the evidence hold up across different diseases?
The evidence is strongest in cancer, but promising in other areas too. The CML study [2] is the most direct example of AI handling patient diversity, but it's limited to one disease. A 2025 study on endometriosis used an AI platform to identify novel drug targets and repurposed drugs, but it focused on a single disease and did not explicitly test across diverse patient subgroups [5]. Similarly, a 2024 study on sclerostin inhibitors for bone disease used AI to build a 3D-QSAR model, but the model was validated on a test set of FDA-approved drugs, not on diverse patient populations [3]. So while the methods are being developed, the evidence for broad cross-population reliability is still thin.
About These Sources
This answer is built on 5 peer-reviewed studies — published from 2022 to 2026, 4 from 2024 or later, 3 in Q1 journals, collectively cited 114 times — selected as the most relevant from 5 studies that passed quality screening, drawn from 59 papers retrieved from a database of over 500 million.
Sources used in this answer
From Algorithms to Assets: A Comprehensive Review of AI's Role in Preclinical Drug Discovery and the Hurdles to Clinical Translation.
This 2026 review identifies patient heterogeneity as a key translational hurdle for AI in drug discovery, alongside data quality and model interpretability, and notes that many models struggle with diverse populations [1].
From biological complexity to clinical precision: A first-in-class single-patient–level AI strategy for precision drug repurposing in relapsed, refractory, or metastatic cancers using COSMIC mutational signatures with blast-crisis CML as a model.
In a 2026 study on blast-crisis CML, an AI strategy using COSMIC mutational signatures generated unique drug repurposing profiles for each of 12 patients, with no two patients sharing the same recommended therapy, demonstrating true N-of-1 precision [2].
AI-based 3D-QSAR model of FDA-approved repurposed drugs for inhibiting sclerostin.
A 2024 study used AI-based 3D-QSAR to model FDA-approved drugs for inhibiting sclerostin, achieving a statistically significant model (q2=0.532, r2=0.969), but the model was validated on a test set of drugs, not on diverse patient populations [3].
Structure-based drug repurposing: Traditional and advanced AI/ML-aided methods
A 2022 review of structure-based drug repurposing discusses traditional and AI methods, noting that AI models often rely on static molecular structures and may not capture dynamic biological differences between patients [4].
Utilizing AI for the Identification and Validation of Novel Therapeutic Targets and Repurposed Drugs for Endometriosis.
A 2025 study used an AI platform to identify novel drug targets (GBP2, HCK) and a repurposed drug (Lifitegrast) for endometriosis, validated in mouse models, but did not explicitly test across diverse patient subgroups [5].