資料載入中.....
|
請使用永久網址來引用或連結此文件:
http://libir.tmu.edu.tw/handle/987654321/64955
|
| 題名: | The Biopsychosocial, Pharmacological, and Non-Pharmacological Factors Associated Depression in Taiwan Population: A Machine Learning, Bioinformatics, and Meta-Analysis Study |
| 作者: | LESMANA, MOH. HENDRA SETIA |
| 貢獻者: | 護理學系博士班
Min-Huey Chung, PhD, RN, FAAN |
| 關鍵詞: | Depression;single-nucleotide polymorphism;biopsychosocial factor;machine learning;random forest;bioinformatics;drug repurposing;functional annotation;genetic;interleukin 6 receptor;non-pharmacological;exercise;mind-body practice;acupuncture;genomic analysis;genomic variants |
| 日期: | 2024-01-18 |
| 上傳時間: | 2025-01-06 |
| 摘要: | Background: Depression represents a significant global burden, posing a substantial challenge in comprehending its determinants and developing effective treatment strategies. This dissertation provided a comprehensive insight into biopsychosocial factors, including genetic factors, and potential pharmacological and non-pharmacological treatments for depression in the Taiwanese population. Objectives: The specific aims of this dissertation were to: (1) explore important biopsychosocial factors, including genetics associated with depression, and generate a predictive model using machine learning analysis. (2) prioritize depression-related gene targets through a functional annotation approach and subsequently identify potential drug repurposing candidates based on these gene targets. (3) test the effectiveness of non-pharmacological treatment on biomarkers related to depression using a meta-analysis approach. Methods: Employing a machine learning approach, we analyzed genome-wide single-nucleotide polymorphism (SNP) data from 4,495 Taiwanese participants genotyped with the Taiwan Biobank chip. Using the synthetic minority oversampling technique to address data imbalance, we selected SNPs through a two-step process using scikit-learn and random forest. Eleven machine learning algorithms were compared to identify the most effective predictive model, and we evaluated models based on SNPs, biopsychosocial factors, and a combination of both. In a bioinformatics approach, we annotated 5.885 SNPs using HaploReg v4.1 and applied five sets of functional annotations to identify depression-associated genes. The list of target genes was expanded using the STRING database, and potential drug candidates were explored in the DrugBank database. Validation involved cross-referencing results with data from the ClinicalTrial.gov and PubMed databases. In a meta-analysis, six databases were searched for relevant articles, and eligibility assessment, data extraction, and analysis were conducted to assess the total effect of non-pharmacological interventions on depression-related biomarkers. Results: Machine learning approach, the random forest (RF) algorithm demonstrated superior performance compared to the other ten algorithms, achieving an accuracy exceeding 92%. The top 30 critical features, encompassing 26 SNPs and four biopsychosocial factors, exhibited exceptional predictive capabilities for depression. The combined SNPs and biopsychosocial model yielded a satisfying prediction matrix. The intronic SNPs in our findings offer potential as depression markers. Bioinformatics approach, we found 7 genes that exhibited a robust association with depression (score=4). Notably, IL6R emerged as a promising candidate, and sarilumab and satralizumab were identified as particularly promising drugs for depression. Meta-analysis approach, non-pharmacological interventions significantly increased neutrophic markers and decreased inflammatory markers. Subgroup analysis of neutrophic markers was significant for mind-body practice and exercise but not significant for nutritional intervention. In subgroup analysis of inflammatory markers, acupuncture, mind-body practice, and exercise were significant. Conclusion: Our findings emphasized the effectiveness and robustness of the RF algorithm in constructing predictive models for depression by leveraging genetic and clinical data. This study reinforced the efficacy of machine learning approaches in discovering novel SNPs associated with depression. Additionally, our results suggested that bioinformatics analysis of SNPs could facilitate the identification of drugs suitable for repurposing in depression treatment, with IL6R identified as a promising gene target, along with Sarilumab and Satralizumab for depression. Furthermore, we proposed the effectiveness of non-pharmacological interventions targeting neurotrophic and inflammatory biomarkers associated with depression, particularly through exercise, mind-body practices, and acupuncture. |
| 描述: | 博士
指導教授：Min-Huey Chung, PhD, RN, FAAN
口試委員：Chien Yu Lai
口試委員：Shiow-Jing Wei
口試委員：Fang-I Hsieh
口試委員：Nguyen Quoc Khanh Le |
| 附註: | 論文公開日期：2029-01-30 |
| 資料類型: | thesis |
| 顯示於類別: | [護理學系] 博碩士論文
|
文件中的檔案:
| 檔案 |
描述 |
大小 | 格式 | 瀏覽次數 |
|---|
| index.html | | 0Kb | HTML | 0 | 檢視/開啟 |
|
在TMUIR中所有的資料項目都受到原著作權保護.
|
