Inferring Correlation Networks from Genomic Survey Data

By Jonathan Friedman, Eric J. Alm. PLoS Comput. Biol., 2012

2,376 citations recorded by the source corpus.

High-throughput sequencing based techniques, such as 16S rRNA gene profiling, have the potential to elucidate the complex inner workings of natural microbial communities - be they from the world's oceans or the human gut. A key step in exploring such data is the identification of dependencies between members of these communities, which is commonly achieved by correlation analysis. However, it has been known since the days of Karl Pearson that the analysis of the type of data generated by such techniques (referred to as compositional data) can produce unreliable results since the observed data take the form of relative fractions of genes or species, rather than their absolute abundances. Using simulated and real data from the Human Microbiome Project, we show that such compositional effects can be widespread and severe: in some real data sets many of the correlations among taxa can be artifactual, and true correlations may even appear with opposite sign. Additionally, we show that community diversity is the key factor that modulates the acuteness of such compositional effects, and develop a new approach, called SparCC (available at https://bitbucket.org/yonatanf/sparcc), which is capable of estimating correlation values from compositional data. To illustrate a potential application of SparCC, we infer a rich ecological network connecting hundreds of interacting species across 18 sites on the human body. Using the SparCC network as a reference, we estimated that the standard approach yields 3 spurious species-species interactions for each true interaction and misses 60% of the true interactions in the human microbiome data, and, as predicted, most of the erroneous links are found in the samples with the lowest diversity.

Subfields: Molecular Biology

View Paper: https://www.semanticscholar.org/paper/538b042c7287ab2159033189da776bbc7b434aae

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Molecular Biology

Citations in the corpus

Effects of diurnal variation of gut microbes and high fat feeding on host circadian clock function and metabolism (2015, neutral): “Detecting significant oscillatory patterns in 16S rRNA amplicon sequencing data is complicated due to lack of absolute OTU abundance information and limited tools to describe nonlinear relationships in relative abundance data (Friedman and Alm, 2012; David et al., 2014).”
Host lifestyle affects human microbiota on daily timescales (2014, methodology): “However, this standard technique leads to what is known as compositional effects [50], and may cause false relationships between OTUs, or between OTUs and metadata.”
Analysis of composition of microbiomes: a novel method for studying microbial composition (2015, background): “Recently, Friedman and Alm (11) introduced the method of sparse compositional correlation (SparCC) to analyze correlation networks among OTUs in 16S rRNA amplicon studies.”
Identifying Keystone Species in the Human Gut Microbiome from Metagenomic Timeseries Using Sparse Linear Regression (2014, positive): “SparCC) to mitigate these problems [14].”
NetCoMi: network construction and comparison for microbiome data in R (2020, positive): “One possible sparsification strategy consists of defining a cut-off value (or threshold) so that only taxa with an absolute association value above this threshold are connected [27, 67].”
Disentangling Interactions in the Microbiome: A Network Perspective (2016, methodology): “SparCC (Sparse Correlations for Compositional data), [5_TD$DIFF]by [6_TD$DIFF]contrast, infers associations in compositional data by estimating the linear Pearson's correlations between the log-transformed components [62].”
Conducting a Microbiome Study (2014, background): “We recommend against this approach, as we have found that it is subject to statistical pitfalls and can lead to samples clustering by sequencing depth (Friedman and Alm, 2012; C. Lozupone, J.G.C., and R.K., unpublished data).”
Deciphering microbial interactions and detecting keystone species with co-occurrence networks (2014, background): “…sulfate reducers that mediate a major biogeochemical process (Pester et al., 2010) and primary degraders of recalcitrant substrates in Frontiers in Microbiology | Microbial Symbioses May 2014 | Volume 5 | Article 219 | 10 the gut that make available substrates for other organisms (Ze et al., 2012).”
Bayesian-multiplicative treatment of count zeros in compositional data sets (2015, methodology): “These important properties are not satisfied by other strategies applied in the existing literature (e.g., Aebischer et al., 1993; Elston et al., 1996; Friedman and Alm, 2012; Rodrigues and Lima, 2009), where the count zero in the vector of proportions is replaced by a small value, but the constant ”
Network analysis methods for studying microbial communities: A mini review (2021, methodology): “Sparse Correlations for Compositional data (SparCC) [26] is a popular method employing this strategy with applications ranging from human gut microbiome studies [27–29] to environmental studies.”