Characterization of the Gut Microbiome in Autism Spectrum Disorders and Atopic Dermatitis

The gut microbiome is broadly recognized as the "second genome" of human beings, that is affected by various factors such as the diet, environment, and the host physiological state. The intestinal microbial composition and structure act a pivotal part in the establishment and the development of the human early immune system and nervous system. With the recognition of the important role of the gut flora in the gut-brain-skin axis, the gut microbiome is considered to have the potential as the therapeutic targets and diagnostic molecular markers for autism spectrum disorders (ASD) and atopic dermatitis (AD). In this thesis, we revealed the characteristics of the gut microbiome of the patients with ASD and AD separately. And we carried out machine learning methods to explore the gut flora molecular markers and stratify the gut bacterial community in eczema infants. Also, we performed the mendelian randomization algorithms to infer the causal relationship between the gut microbiome composition and ASD as well as AD. Firtstly, the eating behavior, diet quality, gastrointestinal symptoms, and gut microbiome composition of a total of 130 children, including 65 children with ASD and 65 typically developing (TD) children were evaluated. We found that the overall gut microbiome composition differed substantially between ASD and TD groups. Compared with the TD children, ASD children showed more eating and feeding problems. ASD children had a lower dietary index, lower dietary diversity, and a lower intake of selected nutrients than TD children. In another microbiome study, 234 fecal samples from AD and healthy adults were collected and sequenced using the 16S rRNA method. The gut microbiome composition, biodiversity, ecological network, and functional profiling were analyzed. Our results illustrated that the species richness and the evenness of the gut microbiome in AD patients were significantly higher than that of healthy people and intense differences in the microbial composition and the network were detected across groups. Romboutsia and Clostridium_sensu_stricto_1 was enriched in the normal group, and Blautia, Butyricicoccus, Lachnoclostridium, Eubacterium_hallii_group, Erysipelatoclostridium, Megasphaera, Oscillibacter, and Flavonifractor were enriched in the AD group. To further validate our findings, raw sequencing data of the stool samples of healthy and eczema infants was collected from the public databases. We first used the DMM clustering algorithm to cluster the metagenomic data of 858 infants. The gut microbiome of the infants under 24 months was divided into four types including Bifidobacterium type1, Bifidobacterium type2, Streptococcus type, and Blautia type. Bifidobacterium type 1 occurred more frequently in infants with eczema than the healthy infants and the number of infants with Blautia type in the eczema group was less than that in the healthy group. Comparing the performance of 5 machine learning classification methods on 458 amplicon sequencing samples, we found that the gradient boosting classification algorithm has the best performance. Notably, the top four bacterial species selected by the gradient boosting classifier were located as the connectors in the microecological interaction network of the gut bacterial community in eczema infants. And then using another independent method, mendelian randomization was employed to discover for causal inference. Our results demonstrated that the increase of the abundance of Eggerthella causes the elevating of the atopic dermatitis risk. However, no significant causal association was detected between the gut microbiome composition and the risk of ASD.In summary, this project revealed the characteristics of the gut microbiome of ASD children and AD patients using various types of data mining methods. The deliverables generated will provide valuable insights for the future treatment and diagnosis of ASD and AD.腸道菌群被稱為人類的"第二基因組"，受到環境、飲食、宿主自身的生理狀態等因素的影響，其組成結構紊亂會對宿主早期的免疫系統以及神經系統的發育造成重大影響。隨著腸道菌群在腦-皮膚-腸軸中的重要功能的揭示，腸道菌群被認為有潛力可以作為自閉癥、特應性皮炎等疾病的治療靶點和診斷的分子標記物。本論文揭示了自閉癥以及特異性皮炎病人的腸道菌群特征，並分別應用監督學習和孟德爾隨機化的算法進行特應性皮炎的腸道菌群分子標記物的挖掘和腸道菌群組成與自閉癥、特應性皮炎等疾病的因果關系推斷。在第二章中，我們對總共130名兒童包括65名自閉癥兒童以及65名正常發育兒童的飲食行為、飲食質量、胃腸道癥狀以及腸道菌群組成進行了評估。結果表明，自閉癥兒童和正常兒童的整體腸道微生物組成存在顯著差異。與正常發育兒童相比，自閉癥兒童表現出更多的飲食問題，他們的飲食指數較低，飲食多樣性較低且所選營養素的攝入較低。在第三章中，我們對特異性皮炎和健康成年人的234個糞便樣本進行擴增子測序，並且從腸道微生物的物種組成種類及豐度、生態多樣性、網絡結構以及物種功能五個方面進行分析比較。我們發現特異性皮炎患者的腸道菌群的物種豐富度和均勻度顯著高於健康人群，而且二者腸道菌群組成及網絡結構存在巨大差異，布勞特式菌屬，丁酸梭菌屬，真桿菌屬，巨球型菌屬，以及顫桿菌克屬在特異性皮炎的病人的腸道菌群中顯著富集，而梭菌屬在健康人群的腸道菌群中顯著富集。第四章中，我們收集了公開數據庫中健康嬰兒和濕疹嬰兒的糞便樣本的測序數據，使用DMM聚類算法對其中858例宏基因組數據進行聚類，我們發現2歲以內的嬰兒腸道菌群可以被分為雙歧桿菌1型、雙歧桿菌2型、鏈球菌型以及布勞特氏菌型四種腸型，其中雙歧桿菌1型在濕疹嬰兒中出現的頻率高於正常嬰兒，而濕疹組中布勞特氏菌型的人數要少於正常組。除此之外我們比較了5種機器學習分類器在458例擴增子測序數據中的表現，我們發現梯度提升分類算法表現最優，並且由梯度提升分類算法挑選出的排名前四的物種在濕疹嬰兒腸道微生態互作網絡中處於各個功能模塊連接點的重要位置。在第五章中，我們利用了兩樣本孟德爾隨機化的方法對腸道菌群物種的組成成分和自閉癥、特異性皮炎以及過敏性疾病進行了因果關系推斷。結果表明，腸道菌群中緩慢艾格式菌的豐度的增加會提高特異性皮炎的患病風險，除此之外，我們還發現，瘤胃菌科，雙歧桿菌，擬桿菌屬的豐度的增加以及費式細菌和普氏梭菌豐度的下降會提高過敏性疾病的患病風險。但是我們沒有發現腸道菌群組成的改變與自閉癥患病風險之間存在因果關聯。總而言之，本論文首先揭示了自閉癥兒童的飲食習慣和腸道菌群的特征以及二者之間的關聯為自閉癥的治療以及飲食幹預提供了有價值的見解。其次，我們揭示了香港成年人特異性皮炎病人腸道菌群的特征，並且利用監督學習的算法基於腸道菌群的數據建立了嬰兒特異性皮炎的分類器，為特異性皮炎的診斷和治療提供了新靶點和新思路。再次我們對腸道菌群組成和疾病進行了因果關系推斷，從系統遺傳學的角度挖掘腸道菌群與疾病之間更深層次的關聯，為疾病的治療提供新的方向。WANG, Yiwei.Ph.D. Chinese University of Hong Kong 2021.Includes bibliographical references (leaves )Abstracts in English and Chinese.Title from PDF title page (viewed on ...