kernInt: A Kernel Framework for Integrating Supervised and Unsupervised Analyses in Spatio-Temporal Metagenomic Datasets

Abstract

The advent of next-generation sequencing technologies allowed relative quantification of microbiome communities and their spatial and temporal variation. In recent years, supervised learning (i.e., prediction of a phenotype of interest) from taxonomic abundances has become increasingly common in the microbiome field. However, a gap exists between supervised and classical unsupervised analyses, based on computing ecological dissimilarities for visualization or clustering. Despite this, both approaches face common challenges, like the compositional nature of next-generation sequencing data or the integration of the spatial and temporal dimensions. Here we propose a kernel framework to place on a common ground the unsupervised and supervised microbiome analyses, including the retrieval of microbial signatures (taxa importances). We define two compositional kernels (Aitchison-RBF and compositional linear) and discuss how to transform non-compositional beta-dissimilarity measures into kernels. Spatial data is integrated with multiple kernel learning, while longitudinal data is evaluated by specific kernels. We illustrate our framework through a single point soil dataset, a human dataset with a spatial component, and a previously unpublished longitudinal dataset concerning pig production. The proposed framework and the case studies are freely available in the kernInt package at https://github.com/elies-ramon/kernInt.

The value of gut microbiota to predict feed efficiency and growth of rabbits under different feeding regimes

“…A second challenge was to define an appropriate way to expand M for those animals in which cecal microbiota was not assessed. These developments are strongly linked with several prediction tools based on kernel methods already proposed 14 . In our study, we have derived kernel matrices by implementing an ad-hoc solution to transform distance matrices into proper covariance matrices, while Ramon et al 14 directly derived the kernel matrices associated with distance metrics from raw information.…”

“…These developments are strongly linked with several prediction tools based on kernel methods already proposed 14 . In our study, we have derived kernel matrices by implementing an ad-hoc solution to transform distance matrices into proper covariance matrices, while Ramon et al 14 directly derived the kernel matrices associated with distance metrics from raw information. Not having microbial information for all the animals under study would request, anyhow, some heuristics to generate valid covariance matrices to be included in the mixed models.…”

Leveraging host-genetics and gut microbiota to determine immunocompetence in pigs

“…Desulfovibrio is a sulfate-reducing bacteria (SRB), which can promote the metabolism of sugars [59] and plays also a key role in intestinal hydrogen and sulfur metabolism [60]. In pigs, Desulfovibrio plays a relevant role during pig gut colonization [49] and was among the dominant genus in healthy pigs compared with diarrhea-affected piglets [61]. In fact, in weaned piglets, a negative correlation between Desulfovibrio and several inflammatory markers such as IL-1β, IL-2 and IL-6, have been observed [62], which would be in agreement with the negative correlation observed between Desulfovibrio and LEU and MON counts in our piglets.…”

The Value of Gut Microbiota to Predict Feed Efficiency and Growth of Rabbits Under Different Feeding Regimes

“…These developments are strongly linked with several prediction tools based on kernel methods already proposed [14-Ramon et al, 2021]. In our study, we have derived kernel matrices by implementing an ad-hoc solution to transform distance matrices into proper covariance matrices, while Ramon et al ( 2021) [14] directly derived the kernel matrices associated with distance metrics from raw information. Not having microbial information for all the animals under study would request, anyhow, some heuristics to generate valid covariance matrices to be included in the mixed models.…”

“…For each trait, the corresponding dataset was randomly divided into 5 folds, 4 of which constituted the learning dataset, and the remaining was used as the validation dataset. Before fitting the sPLSR on the learning dataset, optimal tuning parameters sparsity and number of latent components were chosen by an internal 5-fold cross-validation using cv.spls() function of the "spls" R package [64-Chung et al, 2019] within ranges (0.01-0.99) and (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20) for sparsity and number of latent components, respectively. With the tuning parameters returned by the cv.spls() function, the combination that resulted in the minimum mean squared prediction error (MSPE) was used to finally fit the sPLSR to the learning dataset by the function spls().…”