Foods 2020 , 9 , 1147
3of 19
of circulating cytokines, HDL and LDL concentrations as well as glucose testing into our trial design to assess the e ff ect of rice NPN on these parameters. In this study, we use AI to identify and validate a natural peptide network (NPN), derived from rice, demonstrating anti-inflammatory activity in vitro. Using a machine learning approach, we further characterize rice NPN by identifying a number of constituent bioactive peptides. Finally, to assess associated health benefits of attenuating circulating TNF- α concentrations in an “inflammaging” population, a pilot study was conducted with daily administration of rice NPN over 12 weeks. We show that rice NPN significantly attenuated TNF- α production within 4 weeks of administration. Furthermore, we show significant improvements on oral glucose tolerance testing and a decrease in serum LDL with a concomitant increase in HDL. Finally, in the pilot study, significant physical gains observed, for example, with the chair stand and short physical performance battery tests. In short, an anti-inflammatory rice NPN may have significant benefits in age-related inflammation.
2. Materials and Methods
2.1. Peptide Prediction for Bioactivity and Natural Source Identification To develop a functional peptide hydrolysate that can reduce TNF- α , a predictive machine learning approach was first used to predict peptides with anti-inflammatory activities. To assess this activity, an untargeted approach was applied; structured and unstructured data sources including scientific literature, patents, and public databases were interrogated for anti-inflammatory peptides, with a focus on those with the ability to attenuate TNF- α secretion. For quality assurance, all data were manually curated. A neural network predictive architecture was trained in 10-fold cross validation, once a reliable non redundant dataset of labelled anti-inflammatory peptides was attained (~10 4 data points). The best models for the validation sets were refined on a set of peptides specifically labelled for TNF- α inhibition. This resulted in a smaller set of peptides (~2 × 10 2 ), which included a set of proprietary peptides that had been previously validated for anti-TNF- α activity in-house. The ensemble of the 10 resulting models was used to predict an additional set of proprietary peptides with experimentally determined e ff ect onTNF- α secretion, exhibiting an accuracy higher than 85%. The resulting predictor was finally used to identify bioactive peptides in the proteomes of all the food sources available from our in-house repository. In order to do that, we fragmented the top 10 most abundant proteins of each source into peptides with lengths from 4 to 30 amino acids. Then we scored the resulting set of ~6 × 10 6 peptides using the predictor. We found Oryza sativa as the food source containing the highest number of predicted anti-inflammatory peptides to be unlocked from its most abundant proteins. From these predicted anti-inflammatory peptides, a final set of 8 highly ranked novel peptides were selected for further investigation. All bioactive peptides used in this study were produced by GenScript (Piscataway, NJ, USA), where peptide sequence and purity (95%–99%) were validated by HPLC–MS / MS. 2.2. Natural Peptide Network Production Following the high number of peptides predicted to reduce the expression of TNF- α found in the Oryza sativa L. subsp. japonica proteome, methods similar to Rein et al., 2019, were used to obtain a hydrolysate, rice NPN, which contained these peptides, including the final set of 8 highly ranked peptides [26]. Briefly, we solubilized brown rice protein and sequentially hydrolyzed it with food grade serine protease, in a pH and temperature controlled aqueous solution. The solution was dispersed using a high shear mixer in reverse osmosis water and pH adjusted using sodium hydroxide or hydrochloric acid for a final value of pH 6. The hydrolysis reaction was terminated by heating the solution to 85 ◦ C for 10 min. Immediately after, the solution was rapidly cooled using a plate cooler and stored overnight at 4 ◦ C. The solution was spray dried at 180 ◦ C using a multi-stage Anhydro Spray Dryer at a facility accredited to FSC2200.
Powered by FlippingBook