AI-based target prediction maps the pharmacology of antidiabetic drugs for type 2 diabetes

Researchers have compiled a target-prediction profile of the major drug classes used to treat type 2 diabetes mellitus, using an artificial intelligence tool to estimate which human proteins each antidiabetic agent is most likely to bind. The work, published as a mini review in Discover Medicine, reorganised figures generated by Swiss Target Prediction (STP) for seven groups of antidiabetic agents, from alpha-glucosidase inhibitors to thiazolidinediones, and reports the percentage of main target predictions achieved for each class.

The review, authored by David Calderón Guzmán and colleagues, was motivated in part by the COVID-19 pandemic, which the authors describe as having complicated diabetes management for vulnerable populations. They note that people with diabetes who contracted COVID-19 experienced higher mortality and a higher proportion of critical illness than infected patients without diabetes, a difference the authors associate with the effects of hyperglycemia on systemic inflammatory responses and immune system dysfunction. Age, environmental conditions, socioeconomic status, comorbidities and interventional timing further increase the risks of severe infection, morbidity and mortality in these patients, according to the review.

The methodological core of the study is a ligand-based prediction approach. STP works by comparing a query molecule with the known ligands of a large collection of protein targets, so predictions rest on structural similarity rather than on direct experimental binding measurements. The authors describe the tool as a tuned algorithm with a novel data bank presented through a web interface, and state that the reorganised STP figures allow predictions on therapy combinations and probable side effects through biochemical pathways. They add that the tool helps to understand molecular mechanisms underlying a given phenotype or bioactivity, rationalise possible favourable or unfavourable side effects, predict off-targets of known molecules, and clear the way for drug repurposing.

The first class examined is the alpha-glucosidase inhibitors, currently represented by acarbose and miglitol, with a reported 80 percent main target prediction. These agents act in the small intestine, where they inhibit enzymes such as glucoamylase, sucrase, maltase and isomaltase that convert complex, non-absorbable carbohydrates into simple absorbable carbohydrates, thereby halting carbohydrate absorption. Acarbose features a valienol moiety, an unsaturated aminocyclitol, linked via nitrogen to an isomaltotriose trisaccharide, while miglitol features a piperidine ring with multiple hydroxyl groups and a hydroxyethyl group attached to the nitrogen atom.

The second class is the dipeptidyl peptidase 4 inhibitors, including alogliptin, linagliptin, saxagliptin and sitagliptin, with main target predictions reported between 66.7 and 80 percent. These substances inhibit DPP-4 activity in peripheral plasma; because DPP-4 inactivates the incretin hormone glucagon-like peptide-1, which stimulates insulin secretion, the inhibitors prevent that inactivation and thereby allow GLP-1 to continue stimulating insulin secretion. The review also details the distinctive chemical scaffolds of each agent, from the xanthine-based structure of alogliptin benzoate to the (3R)-3-amino triazolopyrazinyl butanone core of sitagliptin, commonly formulated as a phosphate monohydrate salt.

Meglitinides, represented by nateglinide and repaglinide, form the third class, with 73.3 to 86.6 percent main target prediction. These drugs reduce blood glucose levels by stimulating endogenous insulin production, and the review notes that dosing depends on hemoglobin A1c level, with a recommended starting dose of 0.5 mg pre-prandial for A1c below 8 percent (64 mmol/mol) and 1 to 2 mg for A1c above that threshold. The fourth class, non-sulfonylureas, contains a single currently available agent, metformin, which achieved an 86.6 percent main target prediction. Metformin inhibits the amount of glucose produced by the liver, increases insulin-receptor binding and stimulates tissue uptake of glucose, and consists structurally of a biguanide group with two methyl groups at the 1-position.

Sodium-glucose co-transporter 2 inhibitors, described in the review as the latest class of anti-hyperglycemic agents approved by the US Food and Drug Administration, include canagliflozin, dapagliflozin, empagliflozin and ertugliflozin, with 73.3 to 86.6 percent main target prediction. They reduce renal tubular glucose reabsorption, helping to lower blood glucose without stimulating insulin release. The sulfonylureas, with six agents including glimepiride, glipizide, glyburide, tolazamide, tolbutamide and gliclazide, directly stimulate insulin release from pancreatic beta cells and therefore are useful only in patients with some beta cell function; the review notes that adverse effects may include weight gain and hypoglycemia, and reports 53.3 to 86.6 percent main target prediction for this class. Finally, the thiazolidinediones, rosiglitazone and pioglitazone, increase insulin sensitivity by acting on adipose tissue, muscle and, to a lesser extent, liver, with 59.8 to 66.5 percent main target prediction.

Beyond the target tables, the review situates these drug classes within the broader burden of type 2 diabetes, which it characterises as a chronic metabolic disease with high morbidity and mortality and one of the priority non-communicable diseases worldwide. It describes multifactorial management combining lifestyle changes, such as reduced sodium and fat intake, regular physical activity, weight loss in overweight patients and smoking cessation, with drug therapy, and cites strict glycemic control targets including fasting plasma glucose between 4.4 and 6.1 mmol/L and 2-hour postprandial plasma glucose between 6.1 and 7.8 mmol/L for certain patient groups. The review also notes that monotherapy with a single agent does not generally provide durable glycemic control, so second-line or third-line agents are typically added when side effects appear or glycemic control deteriorates.

The authors devote substantial attention to the challenges of treating older adults, in whom polypharmacy is widespread and has been associated with adverse outcomes including adverse drug events and disability. They report that drug dosages in polypharmacy among the elderly vary according to disease, age, sex and weight, and that drug toxicity in old age is influenced by changes in the quantity, affinity and responsiveness of drug targets, physiological reserve and response to injury. The review also cites cost analyses indicating that, as first-line agents, SGLT2 inhibitors and GLP-1 receptor agonists would improve outcomes in type 2 diabetes but their costs would need to fall by at least 70 percent to be cost-effective, with specific per-person-per-year price targets calculated from published economic evaluations.

The authors conclude that their findings provide real-world evidence supporting the potential clinical benefits of diabetes drug agents in mitigating side effects in patients with type 2 diabetes, while acknowledging that future research could help clarify the many existing uncertainties. Because the work is a mini review based on ligand-similarity predictions and parameters derived from published literature, rather than a new clinical trial, the target percentages and therapeutic recommendations should be read as computational estimates and literature syntheses that would require experimental and clinical validation. The review nevertheless offers a consolidated reference for the molecular targets, chemical structures and predicted off-target profiles of the seven principal antidiabetic drug classes currently in use.

Subject of Research: Artificial intelligence-based prediction of molecular targets of antidiabetic drug classes for type 2 diabetes

Article Title: Antidiabetic drugs and its pharmacological profile target in human for type 2 diabetes mellitus

Article References: Guzmán, D. C., Brizuela, N. O., Peraza, A. V., Olguín, H. J., Herrera, M. O., Jimenez, F. T., del Angel, D. S., & Garcia, E. H. (2026). Antidiabetic drugs and its pharmacological profile target in human for type 2 diabetes mellitus. Discover Medicine, 3(1), Article 157. https://doi.org/10.1007/s44337-026-00659-9

Image Credits: AI Generated

DOI: 10.1007/s44337-026-00659-9

Keywords: type 2 diabetes, antidiabetic drugs, Swiss Target Prediction, artificial intelligence, drug targets, SGLT2 inhibitors, DPP-4 inhibitors, sulfonylureas, drug repurposing, COVID-19, polypharmacy, pharmacology