Bioinformatics investigation of biological characteristics of cell-free DNA to enhance non-invasive prenatal testing and cancer detection

Ph.D.There is increasing research interest in studying the biological characteristics of circulating cell-free DNA (cfDNA) as well as its clinical applications. The discoveries of circulating fetal DNA in the plasma of pregnant women, of donor-specific DNA in patients who have undergone an organ transplantation, and tumor-derived DNA in cancer patients have opened up numerous new possibilities for noninvasive prenatal testing (NIPT), transplant monitoring, and cancer liquid biopsy. In this thesis, I have developed novel bioinformatics approaches with which to assiduously study the biological properties of cell-free DNA. I have then incorporated these novel findings into refined bioinformatics platforms to enhance noninvasive prenatal testing and cancer detection. I have performed detailed investigations of plasma DNA fragment sizes, of preferential cleavage sites of plasma DNA derived from different tissues (i.e., plasma DNA preferred ends), and the relationship between fragment end points and nucleosome structures.In the first section, I have briefly summarized the biology of cfDNA and its current clinical applications.In the second section, I described my work on the development of a novel bioinformatics algorithm with which to determine the genomic locations that are preferentially cleaved in fetal and maternal DNA genomes, referred to as the fetal-preferred and maternal-preferred ends. I have also studied the relationship between plasma DNA preferred ends and nucleosome structures. It was found that fetal DNA was frequently cut within the nucleosomal cores, whereas maternal DNA was cut mostly within the nucleosomal linker regions—discoveries that offered new insights into the mechanism of plasma DNA generation whereby fetal DNA molecules are shorter than maternal DNA molecules. Determination of these preferred ends also allowed us to noninvasively predict fetal DNA fraction while improving the overall performance of NIPT. Fetal DNA fraction is one of the important parameters governing the overall performance of NIPT. Preferred end–based fetal DNA fraction estimation could obviate the need of genotype information and is gender-independent, likely expanding the spectrum of diagnostic applications using cfDNA in maternal plasma.In the third section, I have exploited the known phenomenon of circulating fetal DNA being shorter than circulating maternal DNA. I have studied the fetal DNA fraction in a series of fragment size ranges and identified a characteristic pattern showing the expected change in fetal DNA fraction over the series of fragment size ranges. I then hypothesized that if a chromosomal aneuploidy existed in the fetus, the change in the genomic representation of that chromosome in maternal plasma should also show the expected trend when measured across the series of fragment size windows. This approach is referred to as size-banded copy number aberration analysis. By delineating the patterns of size-banded copy number aberration analysis through the use of a machine learning algorithm, fetal chromosomal aneuploidies can be accurately detected for pregnant women with fetal DNA fraction below 2%. This new approach could potentially lead to improvements of NIPT for pregnant women who have a lower fraction of fetal DNA in maternal plasma. CfDNA-based NIPT is widely applied for detection of fetal chromosomal aneuploidies in millions of pregnant women. However, test failure is not uncommon due to the insufficient amount of fetal DNA molecules in maternal plasma when the fetal DNA fraction is below 4%. In addition, the prevalence of aneuploidy in these pregnant women with fetal DNA fraction below 4% was reported to be 4.7%, which was significantly higher compared with the prevalence of 0.4% in the overall cohort. The new approach developed in this section allowed minimization of test failure caused by low fetal DNA fraction. Size-banded copy number aberration analysis was also demonstrated to greatly enhance the detection of hepatocellular carcinoma (HCC) using plasma DNA.In the fourth section, I have developed a new method to detect human papillomavirus (HPV) integrations in the human genome in a noninvasive manner. Persistent HPV infection can promote the development of head and neck squamous cell carcinoma (HNSCC) and cervical cancer. HPV integration events in the human genome are thought to drive tumorigenesis by increasing genetic instability in the host and dysregulating expression of viral oncogenes. Although some algorithms have been developed for detection of viral integrations, reliable detection of viral integration in plasma DNA remains very challenging, because (1) plasma DNA are fragmented, and most are shorter than 200 bp; (2) the abundance of integrated viral DNA molecules is expected to be very low because the overwhelming background cfDNA are hematopoietic in origin; and (3) unexpected erroneous ligations between human DNA and viral DNA during sequencing library preparation may occur, leading to false positives. The new method described in this section is aimed to overcome these challenges, enabling accurate and specific identification of viral integration sites in tumor tissue and plasma DNA. To do so, it takes advantage of a variety of biological properties of chimeric cfDNA, including those related to fragment sizes and ends.In summary, I have developed a number of bioinformatics approaches that have revealed some novel biological properties of cfDNA molecules and that have enhanced the diagnostic performance of NIPT, cancer detection and viral DNA integration detection.如今，血漿循環遊離DNA的生物學特性和臨床應用研究倍受關註。孕婦血漿中的胎兒遊離DNA，器官移植受體血漿中的供體特異DNA以及癌癥病人血漿中的循環腫瘤DNA的發現為無創產前檢測，器官移植後監測和腫瘤液體活檢的臨床應用開創了諸多可能。在本文中，我開發了創新的生物信息學方法深入研究了血漿遊離DNA的生物學特性，並將這些新的發現整合進了生物信息學平臺以提升無創產前檢測和癌癥檢測的性能。我深入地探索了血漿遊離DNA的片段大小，其在不同組織中的偏好斷裂的基因組位置（如，血漿遊離DNA片段的偏好端點）以及DNA片段的末端位置與核小體結構之間的關系。在本文第一部分，我簡要的總結了血漿遊離DNA的生物學和它目前在臨床應用上的成就。在第二部分，我利用生物信息學算法確定了血漿中胎兒和母體DNA在基因組上的偏好斷裂位置，也可稱為胎兒偏好和母體偏好端點。我還研究了血漿遊離DNA的偏好端點與核小體結構之間的關系，並發現胎兒遊離DNA在核小體的內部區域頻繁斷裂，而母體遊離DNA更多斷裂在核小體間的連接區域，這個發現揭示了在母體血漿中胎兒遊離DNA短於母體遊離DNA的一些機理。鑒定這些偏好端點使我們能夠以非侵入式的方式預測胎兒DNA在母體血漿中的浓度並能提高無創產前檢測的整體性能。胎兒DNA的濃度是影響無創產前檢測效果的重要參數之一。 這種基於偏好端點的胎兒DNA浓度估算方法無需基因型信息，並且與胎兒性別無關，這種方法可能會擴大血漿遊離DNA在母體血漿中的診斷應用範圍。在第三部分中，我探索了在母體血漿中胎兒遊離DNA短於母體遊離DNA的這個現象。我研究了在不同的遊離DNA片段長度範圍內的胎兒DNA濃度並發現一個明顯的現象，即在不同的DNA片段長度範圍內胎兒濃度也會不同。我假設如果胎兒存在染色體非整倍性，那麼在一系列不同長度區間的DNA片段內，母體血漿中該染色體的比例也會顯示出預期的變化趨勢，這個方法也可稱為基於DNA長度區間的拷貝數變異分析。利用機器學習算法結合DNA長度區間的拷貝數變異分析，我們可以精確檢測到胎兒DNA濃度低於2%的孕婦的胎兒染色體非整倍性。這個新方法可以顯著提升無創產前檢測在血漿中胎兒濃度較低的孕婦中的檢測性能。目前，基於血漿遊離DNA的無創產前檢測技術已經被廣泛應用於數百萬孕婦的胎兒染色體非整倍性檢測。但是，當胎兒DNA濃度小於4%時，由於母體血漿中胎兒DNA分子數量不足導致檢測失敗的情況並不少見。此外，據報道，在這些胎兒DNA濃度低於4%的孕婦群體中，胎兒染色體非整倍性的患病率為4.7%，顯著高於整個孕婦群體中0.4%的患病率。本部分中開發的新方法可以降低由於母體血漿中低胎兒DNA濃度導致的無創產前檢測失敗率。同时，我們也證明了這種基於長度區間的拷貝數變異分析方法能夠極大提高利用血漿遊離DNA檢測肝癌的能力。在第四部分，我還開發了一種非侵入式的檢測人體乳頭瘤病毒（HPV）在人類基因組中整合事件的新方法。 HPV的持續感染會促進頭頸部鱗狀細胞癌（HNSCC）和宮頸癌的發生和發展。而人類基因組中的HPV整合事件被認為可以通過增加宿主的遺傳不穩定性和病毒致癌基因的失調表達來驅動腫瘤發生。雖然目前已有一些檢測病毒整合的算法，但要可靠地檢測血漿遊離DNA中的病毒整合事件仍具有挑戰性。因為（1）血漿遊離DNA是片段化的並且大部分的片段長度都小於200bp；（2）由於絕大部分的背景遊離DNA是來自於造血系統，血漿中整合的病毒DNA分子片段豐度預計非常低；（3）測序文庫制備過程中可能會出現人類DNA與病毒DNA的錯誤連接，導致假陽性。本文開發的新方法利用DNA嵌合片段的長度和端點的生物學特性，不僅可以攻克這些挑戰，而且能夠精準特異地鑒定出腫瘤組織或血漿遊離DNA中的病毒整合位點。綜上所述，我開發了一系列的生物信息學方法揭示了血漿遊離DNA分子的一些新的生物學特性，並提升了基於血漿遊離DNA的無創產前檢測，癌癥檢測和病毒整合檢測的性能。Zhang, Haiqiang.Thesis Ph.D. Chinese University of Hong Kong 2019.Includes bibliographical references (leaves 138-147).Abstracts also in Chinese.Title from PDF title page (viewed on 20, January, 2021)