Bioinformatics and Deep Learning Approach to Discover Food-Derived Active Ingredients for Alzheimer’s Disease Therapy

Alzheimer’s disease (AD) prevention is a critical challenge for aging societies, necessitating the exploration of food ingredients and whole foods as potential therapeutic agents. This study aimed to identify natural compounds (NCs) with therapeutic potential in AD using an innovative bioinformatics...

Full description

Saved in:
Bibliographic Details
Main Authors: Junyu Zhou, Chen Li, Yong Kwan Kim, Sunmin Park
Format: Article
Language:English
Published: MDPI AG 2025-01-01
Series:Foods
Subjects:
Online Access:https://www.mdpi.com/2304-8158/14/1/127
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850072123970682880
author Junyu Zhou
Chen Li
Yong Kwan Kim
Sunmin Park
author_facet Junyu Zhou
Chen Li
Yong Kwan Kim
Sunmin Park
author_sort Junyu Zhou
collection DOAJ
description Alzheimer’s disease (AD) prevention is a critical challenge for aging societies, necessitating the exploration of food ingredients and whole foods as potential therapeutic agents. This study aimed to identify natural compounds (NCs) with therapeutic potential in AD using an innovative bioinformatics-integrated deep neural analysis approach, combining computational predictions with molecular docking and in vitro experiments for comprehensive evaluation. We employed the bioinformatics-integrated deep neural analysis of NCs for Disease Discovery (BioDeepNat) application in the data collected from chemical databases. Random forest regression models were utilized to predict the IC<sub>50</sub> (pIC<sub>50</sub>) values of ligands interacting with AD-related target proteins, including acetylcholinesterase (<i>AChE</i>), amyloid precursor protein (<i>APP</i>), beta-secretase 1 (<i>BACE1</i>), microtubule-associated protein tau (<i>MAPT</i>), presenilin-1 (<i>PSEN1</i>), tumor necrosis factor (<i>TNF</i>)<i>-α</i>, and valosin-containing protein (<i>VCP</i>). Their activities were then validated through a molecular docking analysis using Autodock Vina. Predictions by the deep neural analysis identified 166 NCs with potential effects on AD across seven proteins, demonstrating outstanding recall performance. The top five food sources of these predicted compounds were black walnut, safflower, ginger, fig, corn, and pepper. Statistical clustering methodologies segregated the NCs into six well-defined groups, each characterized by convergent structural and chemical signatures. The systematic examination of structure–activity relationships uncovered differential molecular patterns among clusters, illuminating the sophisticated correlation between molecular properties and biological activity. Notably, NCs with high activity, such as astragalin, dihydromyricetin, and coumarin, and medium activity, such as luteolin, showed promising effects in improving cell survival and reducing lipid peroxidation and <i>TNF-α</i> expression levels in PC12 cells treated with lipopolysaccharide. In conclusion, our findings demonstrate the efficacy of combining bioinformatics with deep neural networks to expedite the discovery of previously unidentified food-derived active ingredients (NCs) for AD intervention.
format Article
id doaj-art-ad22ea3a34fc495ea423f4c553ce71b8
institution DOAJ
issn 2304-8158
language English
publishDate 2025-01-01
publisher MDPI AG
record_format Article
series Foods
spelling doaj-art-ad22ea3a34fc495ea423f4c553ce71b82025-08-20T02:47:09ZengMDPI AGFoods2304-81582025-01-0114112710.3390/foods14010127Bioinformatics and Deep Learning Approach to Discover Food-Derived Active Ingredients for Alzheimer’s Disease TherapyJunyu Zhou0Chen Li1Yong Kwan Kim2Sunmin Park3Institute of Advanced Clinical Medicine, Peking University, Beijing 100191, ChinaDepartment of Bioconvergence, Hoseo University, Asan 31499, Republic of KoreaDepartment of Information and Communication Engineering, Hoseo University, Asan 31499, Republic of KoreaDepartment of Bioconvergence, Hoseo University, Asan 31499, Republic of KoreaAlzheimer’s disease (AD) prevention is a critical challenge for aging societies, necessitating the exploration of food ingredients and whole foods as potential therapeutic agents. This study aimed to identify natural compounds (NCs) with therapeutic potential in AD using an innovative bioinformatics-integrated deep neural analysis approach, combining computational predictions with molecular docking and in vitro experiments for comprehensive evaluation. We employed the bioinformatics-integrated deep neural analysis of NCs for Disease Discovery (BioDeepNat) application in the data collected from chemical databases. Random forest regression models were utilized to predict the IC<sub>50</sub> (pIC<sub>50</sub>) values of ligands interacting with AD-related target proteins, including acetylcholinesterase (<i>AChE</i>), amyloid precursor protein (<i>APP</i>), beta-secretase 1 (<i>BACE1</i>), microtubule-associated protein tau (<i>MAPT</i>), presenilin-1 (<i>PSEN1</i>), tumor necrosis factor (<i>TNF</i>)<i>-α</i>, and valosin-containing protein (<i>VCP</i>). Their activities were then validated through a molecular docking analysis using Autodock Vina. Predictions by the deep neural analysis identified 166 NCs with potential effects on AD across seven proteins, demonstrating outstanding recall performance. The top five food sources of these predicted compounds were black walnut, safflower, ginger, fig, corn, and pepper. Statistical clustering methodologies segregated the NCs into six well-defined groups, each characterized by convergent structural and chemical signatures. The systematic examination of structure–activity relationships uncovered differential molecular patterns among clusters, illuminating the sophisticated correlation between molecular properties and biological activity. Notably, NCs with high activity, such as astragalin, dihydromyricetin, and coumarin, and medium activity, such as luteolin, showed promising effects in improving cell survival and reducing lipid peroxidation and <i>TNF-α</i> expression levels in PC12 cells treated with lipopolysaccharide. In conclusion, our findings demonstrate the efficacy of combining bioinformatics with deep neural networks to expedite the discovery of previously unidentified food-derived active ingredients (NCs) for AD intervention.https://www.mdpi.com/2304-8158/14/1/127deep neural analysisneurogenerative diseaseAlzheimer’s diseaseflavonoidsinflammation
spellingShingle Junyu Zhou
Chen Li
Yong Kwan Kim
Sunmin Park
Bioinformatics and Deep Learning Approach to Discover Food-Derived Active Ingredients for Alzheimer’s Disease Therapy
Foods
deep neural analysis
neurogenerative disease
Alzheimer’s disease
flavonoids
inflammation
title Bioinformatics and Deep Learning Approach to Discover Food-Derived Active Ingredients for Alzheimer’s Disease Therapy
title_full Bioinformatics and Deep Learning Approach to Discover Food-Derived Active Ingredients for Alzheimer’s Disease Therapy
title_fullStr Bioinformatics and Deep Learning Approach to Discover Food-Derived Active Ingredients for Alzheimer’s Disease Therapy
title_full_unstemmed Bioinformatics and Deep Learning Approach to Discover Food-Derived Active Ingredients for Alzheimer’s Disease Therapy
title_short Bioinformatics and Deep Learning Approach to Discover Food-Derived Active Ingredients for Alzheimer’s Disease Therapy
title_sort bioinformatics and deep learning approach to discover food derived active ingredients for alzheimer s disease therapy
topic deep neural analysis
neurogenerative disease
Alzheimer’s disease
flavonoids
inflammation
url https://www.mdpi.com/2304-8158/14/1/127
work_keys_str_mv AT junyuzhou bioinformaticsanddeeplearningapproachtodiscoverfoodderivedactiveingredientsforalzheimersdiseasetherapy
AT chenli bioinformaticsanddeeplearningapproachtodiscoverfoodderivedactiveingredientsforalzheimersdiseasetherapy
AT yongkwankim bioinformaticsanddeeplearningapproachtodiscoverfoodderivedactiveingredientsforalzheimersdiseasetherapy
AT sunminpark bioinformaticsanddeeplearningapproachtodiscoverfoodderivedactiveingredientsforalzheimersdiseasetherapy