Machine Learning Discovers a Potent Blood Pressure-Lowering Peptide from Highland Barley

The machine learning pipeline (using Gradient Boosted Decision Trees) successfully predicted ACE-inhibitory peptides from papain-hydrolyzed highland barley protein. The peptide FPRPFL was identified as the most potent ACE inhibitor with an IC50 of 1.18 μM. Enzyme inhibition kinetics, circular dichroism, molecular docking, and molecular dynamics simulations characterized its binding mechanism. Network pharmacology analysis revealed multi-target, multi-pathway antihypertensive properties. The peptide also showed stability in simulated digestion conditions.

Four machine learning models were trained to predict ACE-inhibitory capacity of peptides, with Gradient Boosted Decision Trees (GBDT) performing best. Highland barley protein was hydrolyzed with papain and the resulting peptides were screened computationally. Top candidates were validated through in vitro ACE inhibition assays, enzyme kinetics, and simulated digestion stability tests. Binding mechanisms were analyzed using circular dichroism, molecular docking, and molecular dynamics simulations. Network pharmacology mapped the peptide's multi-target interactions.

Hypertension affects over a billion people worldwide, and many current ACE inhibitors cause side effects like dry cough. Natural food-derived peptides could offer blood pressure management with fewer side effects. This study's combination of AI-driven discovery with thorough lab validation creates a scalable pipeline for finding bioactive peptides — potentially opening the door to functional foods or supplements derived from common crops like barley.

This study exemplifies the convergence of artificial intelligence and bioactive peptide research. Traditional peptide discovery is slow and expensive, but ML-based screening can evaluate thousands of candidates rapidly. As food-derived bioactive peptides gain interest for chronic disease management, this pipeline approach — from grain protein to validated ACE inhibitor — could be applied to many other food sources and therapeutic targets, accelerating the development of peptide-based functional foods and nutraceuticals.

All validation was in vitro — no animal or human studies were conducted. The IC50 value, while impressive in a lab setting, doesn't account for bioavailability, absorption, and metabolism in a living organism. Simulated digestion stability doesn't fully replicate real gastrointestinal conditions. The network pharmacology predictions of multi-target activity are computational and require experimental confirmation. Highland barley protein availability and processing costs for industrial-scale production were not addressed.