利用網路生物學與單細胞轉錄體學探索疾病生物標記與藥物重新定位之應用

Abstract

多體學資料之醫學應用已逐漸普及，許多疾病(例如癌症，新冠肺炎)的預後能受益於體學大數據分析，使病患有更長的存活期或較佳的藥物治療選擇。其中，網絡生物學能運用許多圖的特性來了解生物現象背後的複雜機制。肺癌及肺纖維化是兩個常見且預後不佳的胸腔疾病。隨著藥物的推陳出新，治療的選項也逐漸變多，然而，如何精準的挑選正確的病患族群接受正確的藥物治療，是現階段亟待解決的問題。因此，我們整合了RNA 轉錄體資料，藥物基因體資料以及病患臨床資料，來探討肺癌與COVID-19 肺纖維化 (PCPF) 的生物標記物和新療法。 長鍊非編碼RNA (lncRNA)在肺癌的功能並不完全清楚，我們希望利用基因關聯性網路分析，來探討哪些lncRNA可能會協同作用，並在肺癌中扮演重要角色。我們首先利用TCGA的肺腺癌(TCGA-LUAD)基因表現圖譜(gene expression profile)來構建 lncRNA 關聯網路。並利用關聯性網路中的模組(module)來模擬協同作用的lncRNA。我們發現四個lncRNA 模組表現量顯著地和預後及癌症特徵(cancer hallmarks)相關。我們接著建立由四個模組所組成的基因印記(gene signature)作為一新穎的lncRNA 生物標記。此研究探討了協同運作的lncRNA於肺腺癌的角色，並為日後的lncRNA研究提供了新的契機。 針對長鍊非編碼RNA(lncRNA)當作標靶來治療肺癌的藥物仍十分稀少。我們接著開發一生物網路演算法LncTx來計算藥物目標蛋白和 lncRNA 相關蛋白之間在蛋白質交互作用網路的接近度，期待能找尋靶定與存活相關的lncRNA來達肺癌治療效果的藥物。LncTx比較了不同加權方法在預測藥物適應症的表現，並證實了利用clustering coefficient作為加權權重的效果最佳，我們亦利用了不同的接近度量測方法將現有藥物重新分類，並證實了接近度小的藥物，是較有效的治療藥物。此結果有望在未來做更深入的分析，我們期待lncRNA可能作為新型藥物靶點，並將生物網路接近度之量測方法，作為藥物效度的另一參考。 COVID-19疫情的發展，導致許多感染者產生長久的後遺症，其中包括了肺纖維化(post-COVID-19 pulmonary fibrosis, PCPF)。我們整合了單細胞RNA分析以及許多計算方法來尋找藥物來治療 PCPF。我們發現了一個新的PCPF的基因表達印記，並確認其可作為 PCPF 的治療性生物標記。我們的方法亦找到了數個十分有潛力的治療藥物。我們的研究闡釋了生物網絡的鄰近性可以作為尋找PCPF治療性藥物的框架。 總結來說，網絡科學已被用於分析人類疾病背後的複雜系統。我們利用RNA 轉錄體和藥物基因體數據來探索肺癌和 PCPF 的潛在生物標標記和治療方法。我們的研究結果可作為網路醫學在臨床診斷和治療中的應用的基礎。

The clinical application of omic data is gradually gaining popularity. The prognosis of many diseases (such as cancer and COIVD-19) can benefit from multi-omics analysis; this information provides more precise treatment options which could prolong the survival and bring better life quality. Network biology utilizes the properties of graphs to comprehend the intricate mechanisms underlying biological phenomena. Lung cancer and post-COVID-19 pulmonary fibrosis (PCPF) are two prevailing chest diseases with a poor prognosis. With the introduction of new medications, treatment options are gradually expanding. How to accurately select the correct patient group to receive the correct drug treatment is a pressing issue that must be resolved. Therefore, we integrated RNA transcriptomics data, pharmacogenomic data, and patient clinical data to explore biomarkers and novel therapies for lung cancer and PCPF. The roles of co-regulated long non-coding RNAs (lncRNAs) in lung cancer have not been fully understood. Using the gene association network analysis, we aimed to determine which lncRNAs may co-appear and play a significant role in lung cancer. We first constructed the lncRNA association networks using the gene expression profile of lung adenocarcinoma from The Cancer Genome Atlas (TCGA). We found that the expression of four lncRNA modules was significantly correlated with prognosis and cancer hallmarks. We next established a lncRNA modular signature consisting of four modules as a novel lncRNA prognostic biomarker in lung cancer. This study investigates the role of co-regulated lncRNAs in lung adenocarcinoma and paves the way for lncRNA research in the future. Treatments that target long noncoding RNAs (lncRNAs) in lung cancer are still limited. We next developed a network-based algorithm, LncTx, to calculate the proximity between drug targets and lncRNA-related proteins on the protein-protein interaction network. LncTx compared the performance of various weighting methods for predicting drug indications and showed that the clustering coefficient could be a good weighting parameter. In addition, we reclassified existing drugs using several proximity measures and demonstrated that the drugs with smaller proximity are more effective compared to those with larger proximity. In the future, it is anticipated that the lncRNA may gain its popularity as novel drug targets, and network proximity can be used as an additional reference for the prediction of drug efficacy. As the outbreak of COVID-19, infected patients may suffer from the sequelae of COVID-19, including post-COVID-19 pulmonary fibrosis (PCPF). We integrated single-cell RNA analysis and various computational methods to repurpose drugs for PCPF treatment. We have revealed a novel gene expression signature as a theragnostic biomarker, and identified several potential medications for PCPF. Our study indicates that the single-cell analysis combining network-based proximity may be a novel approach to repurpose medications for PCPF. In conclusion, network science has been utilized to decipher the underlying complicated systems in human disorders. We leveraged single-cell/bulk RNA transcriptomes and pharmacogenomic data to identify potential biomarkers and therapeutics for lung cancer and PCPF. Our findings provide the foundation for the applications of network medicine in clinical diagnosis and therapy.