粒線體與多重器官損傷在COVID-19的相關性之轉錄組和機器學習分析

Abstract

許多研究表明，嚴重急性呼吸道症候群冠狀病毒2型(SARS-CoV-2)可以透過多種方式損害多個器官，包括透過血管收縮素轉化酶2 (ACE2)、促炎性細胞因子風暴或其他次級途徑促進的直接病毒入侵。而long COVID 是指在初次感染 COVID-19 後，個體經歷持續的器官損傷或出現新的症狀。 敗血症引起多重器官損傷的機制包括引發全身性過度發炎反應、免疫系統過度活化、影響細胞能量代謝等幾個重要面向。這些機制相互作用，共同導致多重器官損傷。 本研究利用Gene Expression Omnibus (GEO) 資料庫的公開轉錄組資料來識別在 COVID-19、敗血症和其他急性呼吸道感染疾病中表達顯著差異的基因。進一步的研究檢視與敗血症、其他急性呼吸道感染疾病以及 SARS-CoV-2 誘導的粒線體、心臟、肝臟和腎臟損傷相關的途徑。接著對顯著差異表達的基因進行挑選和排序，並使用特徵重要性對具有生物途徑意義的基因進行排序，作為機器學習驗證的特徵。 透過效能、樣本大小、不平衡資料狀態和過度擬合來評估機器學習的樣本集選擇。機器學習也透過調整基因清單來幫助評估生物途徑的假設。隨後進行的深入研究檢視了基因和相關途徑，以了解粒線體與多重器官損傷在COVID-19 的關聯。 研究結果表明，ACE2、促炎性細胞因子風暴和線粒體損傷對COVID-19 導致的多器官損傷存在著關聯。敗血症和COVID-19引起的多重器官損傷的機制不同。而且，粒線體損傷是導致long COVID的關鍵因素之一。這些發現可以做為研究潛在的醫療治療以應對 SARS-CoV-2 引起的多器官損傷。

The extensive research has shown that Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) can harm several organs through various means, including direct viral invasion facilitated by angiotensin-converting enzyme 2 (ACE2), inflammatory cytokine storms, or other secondary pathways. Long COVID is when individuals experience persistent, lasting damage to organs or new symptoms for several weeks or months after recovering from an initial COVID-19 infection. The mechanisms by which sepsis causes multiple organ damage include several vital aspects, for example, triggering a systemic excessive inflammatory response, the excessive activation of the immune system, and affecting cellular energy metabolism. These mechanisms interact with each other, collectively leading to multiorgan damage. Publicly available transcriptome data from the Gene Expression Omnibus (GEO) database was utilized to identify genes significantly differently expressed in COVID-19, sepsis, and other non-COVID-19 acute respiratory infections. A further investigation examined pathways connected to sepsis, other non-COVID-19 acute respiratory infections, and SARS-CoV-2-induced mitochondrial, cardiac, hepatic, and renal damage. Statistical methods were used to identify and rank significantly differentially expressed genes, and feature importance was used for rating biologically significant genes as machine learning verification features. The sample set selection for machine learning was evaluated through performance, sample size, imbalanced data state, and overfitting. Machine learning also assisted in evaluating biological hypotheses by adjusting gene lists. A subsequently thorough study examined genes and pathways to figure out the correlation between mitochondria and multi-organ damage in COVID-19. The research findings suggest a link between ACE2, inflammatory cytokine storms, and mitochondrial damage in COVID-19, potentially contributing to multiorgan damage. The mechanism of multiorgan damage caused by sepsis and COVID-19 is different. Moreover, mitochondrial damage is one of the critical factors leading to long COVID. These findings indicate that potential medical treatments could be studied to address the damage to multiple organs caused by SARS-CoV-2.