建立用於結核桿菌分析之整合性線上基因型插補平台

Abstract

結核病（TB）由結核分枝桿菌 (M. TB) 所引起，仍然是全球發病和死亡的主要原因之一，因此需要採取有效的控制策略來打破傳播鏈。結核分枝桿菌分成七個主要譜系（譜系1至7），影響結核病的流行病學、傳播及臨床結果。全基因體定序（WGS）和遺傳距離計算對於了解結核病傳播動態已被證明非常有價值。然而，計算遺傳距離的傳統方法在處理缺失的基因型數據時可能會引入偏差。本研究旨在透過基因型插補預測缺失的位點，並將其整合到一個對使用者友善的線上分析平台TBImpact中，從而增強遺傳距離計算。 本研究中收集了來自臺灣高雄市674名細菌學培養呈陽性的肺結核患者的WGS數據。這些數據被用於建立譜系1、2和4的單譜系插補參考模板，包括譜系1、2和4，以及一個混合譜系的模板。利用LinkImputeR，根據序列讀數和LD-kNNi演算法對缺失和低品質基因型進行了插補。TBImpact提供七個主要步驟：序列前處理、譜系鑑定、變異識別、基因型插補、傳播分析、抗藥性預測與系統發育樹生成。此外，TBImpact還提供全流程和逐步分析的選項，為使用者提供了靈活性。透過使用WGS序列檔案作為輸入，TBImpact生成文字格式和HTML格式的結果。 模擬資料集和臨床資料集都被用於評估TBImpact的性能。在模擬資料集中，插補方法表現出高達98.99%的基因型水平精度，超過了傳統方法。從公開的臨床資料集所推測之傳播網路顯示，遺傳距離與原始研究中的結果高度一致，並為結核病傳播的地理距離提供了新的見解。這些發現表明插補方法在恢復結核分枝桿菌基因組資料中的缺失基因型方面具有潛力。 總結來說，本研究提出了一個基於網路的系統，具有結核分枝桿菌基因組的綜合分析流程，整合了基因型插補以提高遺傳距離的準確性，並為研究人員提供了一個高效的結核病分析平台。

Tuberculosis (TB), caused by Mycobacterium tuberculosis (M. TB), remains a major global cause of morbidity and mortality, necessitating effective control strategies to break transmission chains. M.TB has seven major lineages (Lineage 1 to 7), influencing the epidemiology, transmission, and possibly the clinical outcomes of TB. Whole-genome sequencing (WGS) and genetic distance calculations have proven valuable in understanding TB transmission dynamics. Nevertheless, traditional methods for calculating genetic distance may introduce biases when handling missing genotype data. This study aimed to enhance genetic distance calculations by predicting missing sites through genotype imputation and integrating this into a user-friendly online analysis platform, TBImpact. In this study, WGS data were collected from 674 newly diagnosed pulmonary TB patients in Kaohsiung, Taiwan, with positive bacterial cultures. These data were used to establish single-lineage imputation reference panels for Lineages 1, 2, and 4, as well as a mixed lineage panel. Utilizing LinkImputeR, missing and low-quality genotypes were imputed based on the read count and the LD-kNNi algorithm. TBImpact serves seven major steps: sequence preprocessing, lineage specification, variant calling, genotype imputation, transmission analysis, drug resistance prediction, and phylogenetic tree generation. In addition, TBImpact offers both full-pipeline and step-by-step analysis options, providing users with flexibility. Using WGS sequence files as input, TBImpact generates text-formatted and HTML-formatted results. Both simulated datasets and clinical datasets were used to evaluate the performance of TBImpact. In the case of the simulation datasets, the imputation method exhibited an impressive genotype level accuracy of 98.99%, surpassing conventional methods. The inferred transmission network from clinical datasets in public suggested that the genetic distances were closely aligned with those in the original research studies and provided new insights into the geographic distances of tuberculosis transmission. These findings indicated the potential of the imputation method to accurately recover missing genotypes in M.TB genomic data. In conclusion, this study proposed a web-based system featuring a comprehensive analysis pipeline for the M. TB genomes, integrating genotype imputation to improve genetic distance accuracy, and providing researchers with a highly effective analysis platform for TB.