東華大學圖書館 |

Signatures of Health and Disease in the Human Microbiome.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Signatures of Health and Disease in the Human Microbiome./
作者:	Khan, Saad M.
面頁冊數:	1 online resource (327 pages)
附註:	Source: Dissertations Abstracts International, Volume: 84-05, Section: B.
Contained By:	Dissertations Abstracts International84-05B.
標題:	Systematic biology. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=29395104click for full text (PQDT)
ISBN:	9798351480299

Signatures of Health and Disease in the Human Microbiome.
Khan, Saad M.

Signatures of Health and Disease in the Human Microbiome. - 1 online resource (327 pages)

Source: Dissertations Abstracts International, Volume: 84-05, Section: B.

Thesis (Ph.D.)--Albert Einstein College of Medicine, 2023.

Includes bibliographical references

Every person carries within themselves a diverse ecosystem of microorganisms known as their microbiome. The intestinal (gut) microbiome in particular interacts with the host immune system and is thought to play a role in many diseases such as colorectal cancer, inflammatory bowel and diabetes. Although microbiome studies have generated terabytes of data, there are also many issues with reproducibility of associations and consistency of results between studies. Basic research on the microbiome has yet to be widely translated into the clinic.Many of these issues are due to batch effects, uncontrolled variations between studies, and limited sample sizes. However, a larger issue has gone under the radar - many biomarkers identified by studies of different diseases are the same. What does it mean when a paper on the microbiome in diabetes identifies markers that overlap with those from a paper on colorectal cancer? Naturally, one may interpret these findings as evidence that there are common inflammatory processes that have similar effects on a microbiome, but much of the literature ignores this observation when analyzing results. If most of the markers identified for a disease are actually general markers, then how do we find taxa that are specifically associated with that condition and its pathophysiology?Here, we demonstrate that there exists a common set of taxonomic markers that underlie a wide range of diseases that interact with the gut microbiome. These common markers are both a limitation to traditional GWAS-style approaches to studying the microbiome and a useful, novel, framework for thinking about potential mechanisms of host-bacteria interactions in health and disease.Launching this project required substantial computing power and data engineering - individual microbiome samples generate gigabases of genomic reads, and there are thousands such samples and associated metadata. We first present Fireworks, an open source library for reproducible machine learning, that we developed specifically for large scale bioinformatics workloads. Using Fireworks, we build a cloud-based pipeline for performing statistical analysis on dozens of existing metagenomic studies.We then use this foundation to develop machine learning models to classify diseases from annotated metagenomics samples. Using 5643 aggregated, annotated whole-community metagenomes we implement the first multiclass microbiome disease classifier of this scale, able to discriminate between 18 different diseases and healthy. We develop a novel graph convolutional neural network algorithm which provides state-of-the-art classification accuracy, correctly classifying 75% of test samples and achieving a 92.1% overall AUROC. Such a model may have useful diagnostic utility in a clinical setting, although deep learning methods are often criticized for lacking interpretability. Why does the model make certain predictions and therefore what components of a microbiome are predictive for disease? To answer these questions, we turn to more traditional statistical approaches to identify common signatures of health by itself. We develop a novel meta-analysis methodology to assign a score to individual taxa describing their association with health, utilizing 21 different studies and over 6000 samples. Using this methodology, we identify 127 taxa that consistently underlie health and disease and reveal new facets of the biology and potential mechanisms behind microbiome/disease associations.We show a consistent and significant correlation between the oral microbiome and disease. This concept has been discussed previously, but our work is the first to generalize and statistically show that oral microbes in the gut are a consistent signature of disease across datasets. More importantly, we identify specific native oral taxa, such as Gemella morbillorum, which are reproducibly associated with disease in the gut.Our results suggest that the concept of statistical association in the microbiome should be thought of in relative terms. For example, previous meta-analyses have argued that F. nucleatum is a strong marker for colorectal cancer. However, we show that it is also associated with many unrelated diseases, making it a more general marker. We present for the first time a set of taxa that can be thought of as general biomarkers of health and disease, thereby yielding a statistical characterization of the widely used term "dysbiosis".Together, this thesis builds a general foundation for future works which seek to understand the mechanisms of microbiome associations with diseases in the gut. Moreover, our work suggests that the results of the many case-control studies in the literature can be reinterpreted to account for the relative degree of specificity of a particular association, potentially improving microbiome diagnostics and informing efforts to engineer the microbiome to improve patient outcomes.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2023

Mode of access: World Wide Web

ISBN: 9798351480299Subjects--Topical Terms:

3173492
Systematic biology.
Subjects--Index Terms:

BiomarkersIndex Terms--Genre/Form:

542853
Electronic books.

Signatures of Health and Disease in the Human Microbiome.
LDR:06266nmm a2200397K 4500 001 2362784
005 20231109093725.5
006 m o d
007 cr mn ---uuuuu
008 241011s2023 xx obm 000 0 eng d
020 $a 9798351480299
035 $a (MiAaPQ)AAI29395104
035 $a AAI29395104
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Khan, Saad M. $3 3703523
245 1 0 $a Signatures of Health and Disease in the Human Microbiome.
264 0 $c 2023
300 $a 1 online resource (327 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 84-05, Section: B.
500 $a Advisor: Kelly, Libusha.
502 $a Thesis (Ph.D.)--Albert Einstein College of Medicine, 2023.
504 $a Includes bibliographical references
520 $a Every person carries within themselves a diverse ecosystem of microorganisms known as their microbiome. The intestinal (gut) microbiome in particular interacts with the host immune system and is thought to play a role in many diseases such as colorectal cancer, inflammatory bowel and diabetes. Although microbiome studies have generated terabytes of data, there are also many issues with reproducibility of associations and consistency of results between studies. Basic research on the microbiome has yet to be widely translated into the clinic.Many of these issues are due to batch effects, uncontrolled variations between studies, and limited sample sizes. However, a larger issue has gone under the radar - many biomarkers identified by studies of different diseases are the same. What does it mean when a paper on the microbiome in diabetes identifies markers that overlap with those from a paper on colorectal cancer? Naturally, one may interpret these findings as evidence that there are common inflammatory processes that have similar effects on a microbiome, but much of the literature ignores this observation when analyzing results. If most of the markers identified for a disease are actually general markers, then how do we find taxa that are specifically associated with that condition and its pathophysiology?Here, we demonstrate that there exists a common set of taxonomic markers that underlie a wide range of diseases that interact with the gut microbiome. These common markers are both a limitation to traditional GWAS-style approaches to studying the microbiome and a useful, novel, framework for thinking about potential mechanisms of host-bacteria interactions in health and disease.Launching this project required substantial computing power and data engineering - individual microbiome samples generate gigabases of genomic reads, and there are thousands such samples and associated metadata. We first present Fireworks, an open source library for reproducible machine learning, that we developed specifically for large scale bioinformatics workloads. Using Fireworks, we build a cloud-based pipeline for performing statistical analysis on dozens of existing metagenomic studies.We then use this foundation to develop machine learning models to classify diseases from annotated metagenomics samples. Using 5643 aggregated, annotated whole-community metagenomes we implement the first multiclass microbiome disease classifier of this scale, able to discriminate between 18 different diseases and healthy. We develop a novel graph convolutional neural network algorithm which provides state-of-the-art classification accuracy, correctly classifying 75% of test samples and achieving a 92.1% overall AUROC. Such a model may have useful diagnostic utility in a clinical setting, although deep learning methods are often criticized for lacking interpretability. Why does the model make certain predictions and therefore what components of a microbiome are predictive for disease? To answer these questions, we turn to more traditional statistical approaches to identify common signatures of health by itself. We develop a novel meta-analysis methodology to assign a score to individual taxa describing their association with health, utilizing 21 different studies and over 6000 samples. Using this methodology, we identify 127 taxa that consistently underlie health and disease and reveal new facets of the biology and potential mechanisms behind microbiome/disease associations.We show a consistent and significant correlation between the oral microbiome and disease. This concept has been discussed previously, but our work is the first to generalize and statistically show that oral microbes in the gut are a consistent signature of disease across datasets. More importantly, we identify specific native oral taxa, such as Gemella morbillorum, which are reproducibly associated with disease in the gut.Our results suggest that the concept of statistical association in the microbiome should be thought of in relative terms. For example, previous meta-analyses have argued that F. nucleatum is a strong marker for colorectal cancer. However, we show that it is also associated with many unrelated diseases, making it a more general marker. We present for the first time a set of taxa that can be thought of as general biomarkers of health and disease, thereby yielding a statistical characterization of the widely used term "dysbiosis".Together, this thesis builds a general foundation for future works which seek to understand the mechanisms of microbiome associations with diseases in the gut. Moreover, our work suggests that the results of the many case-control studies in the literature can be reinterpreted to account for the relative degree of specificity of a particular association, potentially improving microbiome diagnostics and informing efforts to engineer the microbiome to improve patient outcomes.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2023
538 $a Mode of access: World Wide Web
650 4 $a Systematic biology. $3 3173492
650 4 $a Microbiology. $3 536250
650 4 $a Health sciences. $3 3168359
650 4 $a Pathology. $3 643180
653 $a Biomarkers
653 $a Health
653 $a Meta-analysis
653 $a Microbiome
655 7 $a Electronic books. $2 lcsh $3 542853
690 $a 0423
690 $a 0566
690 $a 0410
690 $a 0571
710 2 $a ProQuest Information and Learning Co. $3 783688
710 2 $a Albert Einstein College of Medicine. $b Biomedical Science. $3 3436140
773 0 $t Dissertations Abstracts International $g 84-05B.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=29395104 $z click for full text (PQDT)