Integrated Metagenomics Analysis Identifies Loss of Diversity in - - PowerPoint PPT Presentation

Integrated Metagenomics Analysis Identifies Loss of Diversity in Periodontitis

Li Charlie Xia Dongmei Ai Medical Oncology / Stanford University Applied Mathematics / University of Science & Technology Beijing GIW 2016

OVERVIEW

Oral microbiome contributes to Periodontitis (PD)

Periodontitis involves progressive loss of alveolar bone around teeth and leads to loss of teeth. Periodontitis affects more than half of world's adult population. Its etiology is unknown, however, oral microbiome (dental plaque) is believed contributor to the disease.

Microbiome is best studied by Metagenomics

Before metagenomics:

• Bacterial isolation and culture
• Biased, tedious and low throughput

Shotgun metagenomics:

• No isolation and culture needed
• High-throughput, low cost
• Less biased, full species spectrum

Integrate metagenomic datasets:

• Effectively increase sample size
• Higher power for statistical analysis

ANALYSIS

Integrate Published Oral Metagenomics Datasets

Debates on Periodontitis:

• One or multiple etiology path?
• One or multiple pathogens?
• Stable or progressing?

Published Periodontitis Metagenomics Datasets:

• 13 subgingival plaque samples (Duran-

Pinedo et al ISME J 2014)

• 30 subgingival plaque samples (Yost et al

Genome Medicine 2015)

13 samples (Duran- Pinedo et al) 30 samples (Yost et al) 43 samples: 6 healthy controls 14 stable periodontitis 16 progressing periodontitis 7 unknown periodontitis

• 1. Input Integrated Metagenomics Datasets

to the Pipeline

• 2. Quality Control and Preprocessing:

§ TagCleaner, PRINSEQ, DeconSeq and FLASH

• 3. Expanded Phylogenetic Analysis:

§ MetaPhylan

• 4. Refined Phylogenetic Analysis

§ BWA-MEM § GRAMMy

• 5. Differential Abundance Analysis
• 6. Logistic Regression Analysis
• 7. Bi-clustering Analysis
• 8. Network Analysis

§ ELSA

Bioinformatics and Statistical Analysis

1 2 3 4 5 6 7 8

GRAMMy: Accurately Finds Microbiome Abundance

Xia et al. PLOS ONE 2011 https://bitbucket.org/charade/grammy/overview

Accurately estimates relative abundance using probabilistic mixture modeling. Allows ambiguous assignments in read mapping. Uses Expectation- Maximization Algorithm to find the MLE of mixing parameter/relative abundance.

ELSA: Efficient Co-occurrence Network Analysis

Xia et al. BMC Systems Biology 2011 Xia et al. Bioinformatics 2013 https://bitbucket.org/charade/elsa

RESULTS

Periodontitis Oral Microbiome

Bar plots of top 20 abundant species stratified by sample disease states. Error bars shows considerable variation among samples even with the same disease state. Many species are found to be top abundant in all states: healthy, stable and progress

• S. gondornii, S. Sanguinis, F. Alocis ...

Some species are found to be top abundant in only diseased states:

• L. gasseri, S. Epidermidis, ...

healthy (green) stable (yellow) progressing (red)

Loss of Microbiome Diversity in Periodontitis

!"

! !!! = −4.343! + 10.212,

Alpha-diversity accurately predicts Periodontitis:

• Shannon Index of 2 seems to be a significant diversity

threshold distinguishing diseased and healthy samples.

• Our logistic model shows d=2 gives p>82%.
• Our Naïve (yes if p>50%) logistic classifier achieves

accuracy 94.4% when applied to new samples not used in training. Periodontitis is associated with Alpha-diversity (*P<0.011): d: alpha-diversity p: probability of Periodontitis

Marker Species in Periodontitis Mcirobiome

A table of plots of single molecule statistics

Species loss of abundance in Periodontitis samples:

• G. Morbillorum, V. Parvula, H. parainfluenzae, C. Matruchotii, N. Flavescens
• L. gasseri gain abundance in progressing Periodontitis samples:

Keynote and Marker Species of Periodontitis

• 1. Marker Species

(Blue): Significantly different abundance between disease states

• 2. New Keystone

Species (Pink): H. haemolyticus, P. Prevotella, C.

• chracea, who share

the same composition profile with known keystone species: Porphyromonas gingivalis.

Microbiome Correlation Networks of Periodontitis

• 1. Significant reduction of edges in diseased sample network.
• 2. Loss of all negative correlation in diseased samples.
• 3. Potential loss of check-and-balance through negative feedbacks.

Keystone Species Mediated Loss of Diversity Model

SUMMARY

Summary We integrated two published Periodontitis metagenomics datasets. We analyzed the integrated datasets with standardized bioinformatics pipelines. We found loss of oral microbiota diversity Is strongly associated with Periodontitis. Alpha diversity accurately predicts Periodontitis disease states. Loss of diversity has been associated with other conditions including obesity. Our finding suggest diversity could be useful measure for early Periodontitis diagnosis and therapeutic intervention.

Bioinformatics Tools for Meta- / Human Genomics

Phylogenetic Abundance Analysis: GRAMMy: https://bitbucket.org/charade/grammy (Xia et al Plos One 2011) Correlation/Network Analysis: ELSA: https://bitbucket.org/charade/elsa (Xia et al Bioinformatics 2013, benchmarked in Weiss et al ISME J 2016) Structural Variant Analysis: SWAN: https://bitbucket.org/charade/swan (Xia et al NAR 2016)

Acknowledgement

Stanford University

• Hanlee Ji
• David Siegmund

University of Science and Technology Beijing

• Ruocheng Huang
• Chao Li
• Jiangping Zhu

University of Pennsylvania

• Nancy Zhang
• Gerard Schellenberg
• Li-san Wang

Funding Source NIH/NHGRI: 2R01HG006137 CNSF: 61370131

Shanghai Jiao Tong University

• Jin Wen

University of Southern California

• Fengzhu Sun
• Jed Fuhrman