Graduate Thesis Or Dissertation
 

Microbial Network Recovery by Compositional Graphical Lasso Under Additive Log-Ratio Transformation

Pubblico Deposited

Contenuto scaricabile

Scarica il pdf
https://ir.library.oregonstate.edu/concern/graduate_thesis_or_dissertations/dn39x784d

Descriptions

Attribute NameValues
Creator
Abstract
  • The interactions between microbial taxa have been under great research interest in the sci- ence community given the microbiome data deluge. Several methods have been proposed to model and estimate the conditional dependency between microbial taxa for their interac- tions, in order to eliminate spurious correlation detections. However, these methods either do not account for the compositional count nature of microbiome data (such as graphi- cal lasso), or are built upon the central log-ratio transformation (such as SPIEC-EASI) that results in a degenerate covariance matrix and thus an undefined precision matrix to present the underlying network. In addition, most existing methods ignore the potential consequence of the heterogeneity nature of microbiome data that the sum of the counts within each sample, termed “sequencing depth”, can vary drastically across samples. To address these issues, we propose a novel method called “compositional graphical lasso” to identify the microbial interactions by adopting a logistic normal multinomial model which explicitly incorporates the sequencing depths. Different from most existing meth- ods, compositional graphical lasso is based on the additive log-ratio transformation, which first selects a reference taxon and then computes the log ratios of the abundances of all the other taxa with respect to that of the reference. One natural concern about the additive log-ratio transformation would be whether the estimated network is invariant with respect to the choice of the reference. To further address this concern, we establish the reference- invariance property of a subnetwork of interest based on the additive log-ratio transformed data, and propose a reference-invariant version of the compositional graphical lasso by modifying the penalty term in its objective function to penalize only the invariant subnet- work. We illustrate the advantages of the proposed methods over the existing ones under a variety of simulation scenarios and also demonstrate their efficacy by applying them to an oceanic microbiome data set.
Contributor
License
Resource Type
Date Issued
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Advisor
Committee Member
Academic Affiliation
Dichiarazione dei diritti
Funding Statement (additional comments about funding)
  • Our research is sponsored by NIH funding R01GM126549
Publisher
Peer Reviewed
Language
Embargo reason
  • Pending Publication
Embargo date range
  • 2020-06-18 to 2022-07-19

Le relazioni

Parents:

This work has no parents.

In Collection:

Elementi

Thumbnail Titolo Data caricata Visibilità Azioni
file details TianChuan2020.pdf 2020-06-18 Pubblico Scaricare