9654467111 Call Girls In Raj Nagar Delhi Short 1500 Night 6000
Using Dimensionality Reduction to Identify Key Species in Polymicrobial Disease: CF as a Case Study
1. Abstract
Identifying the role of specific bacterial species in chronic, polymicrobial infections is key
to developing effective therapeutic strategies. However, the complex ecology of
multispecies infections presents many challenges and requires diagnostics that capture
rare and/or non-standard pathogens. While next-gen sequencing allows the
identification of microbiome composition, selecting informative taxa to monitor or study
in vitro from these high dimensional datasets presents an additional analytical
challenge. We use cystic fibrosis airway microbiomes as a model system to examine
strategies to identify key taxa predictive of disease severity (measured by lung function).
Specifically we focus on a cohort of 77 clinically stable CF patients for which we gather
16S microbiome data plus patient meta-data. To address the challenges of feature
selection on large, zero-inflated datasets, we assess various transformation and dimension
reduction techniques prior to machine learning analyses (LASSO, ElasticNet, or Ridge
Regression). We recover well-studied (pathogens are negative predictors) as well as novel
(putative probiotic taxa) bacteria-host associations. In addition, we identify bacterial
associations dependent on health context (e.g. CF-related diabetic status). To build
towards clinical use, we are extending our data analysis to incorporate published data and
ongoing longitudinal studies in collaboration with the Atlanta CF clinics.
Link microbiome composition to patient health.
Identify biomarkers of lung function.
Identifying Microbial Predictors of Health in People With CF
Conan Zhao Yifei Wang John Varga Joanna Goldberg Arlene Stecenko Sam Brown
1School of Biology, Center for Microbial Dynamics and Infection, Georgia Institute of Technology
2Division of Pulmonary Medicine, Department of Pediatrics, Emory University School of Medicine
3Emory-Children's Center for Cystic Fibrosis Research, Emory University and Children's Healthcare of Atlanta
06-23-2019
MBP-19
Georgia Tech
804.658.7427
czhao98@gatech.edu
Objectives
Methods Overview
Are non-pathogenic taxa informative of patient health?
Primary Findings
Machine Learning Details
Acknowledgements
Funding. Center for Cystic Fibrosis and Airways Disease
Research, and Children’s Healthcare of Atlanta pilot grant
to Drs. Stecenko and Goldberg (initial 77 person cohort).
CDC award No. BAA 2016-N-17812 and BAA 2017-
OADS-01 to Drs. Brown, Stecenko, and Goldberg
IRB. Emory IRB00042577 / GT IRB H18431
Sputum samples were obtained from the CF-BR at
Children's Healthcare of Atlanta and Emory University
Pediatric CF Discovery Core
We would especially like to thank the study participants
for their time and donations.
Longitudinal sampling
Identify the mechanistic role of Fusobacterium
Predicting lung function decline and clinical stability
Elastic Net
70:30 split
Leave-one-out
cross-
validation
1000-fold
Bootstrap
resampling
Severe (<40%) Moderate (40-60%) Mild (60-80%) Normal (>80%)
Other Results
Future Directions
Inclusion of non-pathogens
greatly improves predictive
models of lung function in
people with cystic fibrosis.
Take a picture for:
Full Poster
BioRXiv manuscript
Non-pathogen 16S decreases error
% oral taxa is a positive predictor
Fusobacterium and Rothia as
potential biomarkers
Predictor selection is robust
Thank you for visiting!