Browsing by Author "Tikhonov, Gleb"

Now showing 1 - 6 of 6

Accounting for environmental variation in co-occurrence modelling reveals the importance of positive interactions in root-associated fungal communities
(2020-07-01) Abrego, Nerea; Roslin, Tomas; Huotari, Tea; Tack, Ayco J.M.; Lindahl, Björn D.; Tikhonov, Gleb; Somervuo, Panu; Schmidt, Niels Martin; Ovaskainen, Otso
A1 Alkuperäisartikkeli tieteellisessä aikakauslehdessä
Understanding the role of interspecific interactions in shaping ecological communities is one of the central goals in community ecology. In fungal communities, measuring interspecific interactions directly is challenging because these communities are composed of large numbers of species, many of which are unculturable. An indirect way of assessing the role of interspecific interactions in determining community structure is to identify the species co-occurrences that are not constrained by environmental conditions. In this study, we investigated co-occurrences among root-associated fungi, asking whether fungi co-occur more or less strongly than expected based on the environmental conditions and the host plant species examined. We generated molecular data on root-associated fungi of five plant species evenly sampled along an elevational gradient at a high arctic site. We analysed the data using a joint species distribution modelling approach that allowed us to identify those co-occurrences that could be explained by the environmental conditions and the host plant species, as well as those co-occurrences that remained unexplained and thus more probably reflect interactive associations. Our results indicate that not only negative but also positive interactions play an important role in shaping microbial communities in arctic plant roots. In particular, we found that mycorrhizal fungi are especially prone to positively co-occur with other fungal species. Our results bring new understanding to the structure of arctic interaction networks by suggesting that interactions among root-associated fungi are predominantly positive.
Computationally efficient joint species distribution modeling of big spatial data
(2020-02-01) Tikhonov, Gleb; Duan, Li; Abrego, Nerea; Newell, Graeme; White, Matt; Dunson, David; Ovaskainen, Otso
A1 Alkuperäisartikkeli tieteellisessä aikakauslehdessä
The ongoing global change and the increased interest in macroecological processes call for the analysis of spatially extensive data on species communities to understand and forecast distributional changes of biodiversity. Recently developed joint species distribution models can deal with numerous species efficiently, while explicitly accounting for spatial structure in the data. However, their applicability is generally limited to relatively small spatial data sets because of their severe computational scaling as the number of spatial locations increases. In this work, we propose a practical alleviation of this scalability constraint for joint species modeling by exploiting two spatial-statistics techniques that facilitate the analysis of large spatial data sets: Gaussian predictive process and nearest-neighbor Gaussian process. We devised an efficient Gibbs posterior sampling algorithm for Bayesian model fitting that allows us to analyze community data sets consisting of hundreds of species sampled from up to hundreds of thousands of spatial units. The performance of these methods is demonstrated using an extensive plant data set of 30,955 spatial units as a case study. We provide an implementation of the presented methods as an extension to the hierarchical modeling of species communities framework.
Higher host plant specialization of root-associated endophytes than mycorrhizal fungi along an arctic elevational gradient
(2020-08-01) Abrego, Nerea; Huotari, Tea; Tack, Ayco J.M.; Lindahl, Björn D.; Tikhonov, Gleb; Somervuo, Panu; Martin Schmidt, Niels; Ovaskainen, Otso; Roslin, Tomas
A1 Alkuperäisartikkeli tieteellisessä aikakauslehdessä
How community-level specialization differs among groups of organisms, and changes along environmental gradients, is fundamental to understanding the mechanisms influencing ecological communities. In this paper, we investigate the specialization of root-associated fungi for plant species, asking whether the level of specialization varies with elevation. For this, we applied DNA barcoding based on the ITS region to root samples of five plant species equivalently sampled along an elevational gradient at a high arctic site. To assess whether the level of specialization changed with elevation and whether the observed patterns varied between mycorrhizal and endophytic fungi, we applied a joint species distribution modeling approach. Our results show that host plant specialization is not environmentally constrained in arctic root-associated fungal communities, since there was no evidence for changing specialization with elevation, even if the composition of root-associated fungal communities changed substantially. However, the level of specialization for particular plant species differed among fungal groups, root-associated endophytic fungal communities being highly specialized on particular host species, and mycorrhizal fungi showing almost no signs of specialization. Our results suggest that plant identity affects associated mycorrhizal and endophytic fungi differently, highlighting the need of considering both endophytic and mycorrhizal fungi when studying specialization in root-associated fungal communities.
Joint species distribution modelling with the R-package Hmsc
(2020-03-01) Tikhonov, Gleb; Opedal, Øystein H.; Abrego, Nerea; Lehikoinen, Aleksi; de Jonge, Melinda M.J.; Oksanen, Jari; Ovaskainen, Otso
A1 Alkuperäisartikkeli tieteellisessä aikakauslehdessä
Joint Species Distribution Modelling (JSDM) is becoming an increasingly popular statistical method for analysing data in community ecology. Hierarchical Modelling of Species Communities (HMSC) is a general and flexible framework for fitting JSDMs. HMSC allows the integration of community ecology data with data on environmental covariates, species traits, phylogenetic relationships and the spatio-temporal context of the study, providing predictive insights into community assembly processes from non-manipulative observational data of species communities. The full range of functionality of HMSC has remained restricted to Matlab users only. To make HMSC accessible to the wider community of ecologists, we introduce Hmsc 3.0, a user-friendly r implementation. We illustrate the use of the package by applying Hmsc 3.0 to a range of case studies on real and simulated data. The real data consist of bird counts in a spatio-temporally structured dataset, environmental covariates, species traits and phylogenetic relationships. Vignettes on simulated data involve single-species models, models of small communities, models of large species communities and models for large spatial data. We demonstrate the estimation of species responses to environmental covariates and how these depend on species traits, as well as the estimation of residual species associations. We demonstrate how to construct and fit models with different types of random effects, how to examine MCMC convergence, how to examine the explanatory and predictive powers of the models, how to assess parameter estimates and how to make predictions. We further demonstrate how Hmsc 3.0 can be applied to normally distributed data, count data and presence–absence data. The package, along with the extended vignettes, makes JSDM fitting and post-processing easily accessible to ecologists familiar with r.
Learning conditional variational autoencoders with missing covariates
(2024-03) Ramchandran, Siddharth; Tikhonov, Gleb; Lönnroth, Otto; Tiikkainen, Pekka; Lähdesmäki, Harri
A1 Alkuperäisartikkeli tieteellisessä aikakauslehdessä
Conditional variational autoencoders (CVAEs) are versatile deep latent variable models that extend the standard VAE framework by conditioning the generative model with auxiliary covariates. The original CVAE model assumes that the data samples are independent, whereas more recent conditional VAE models, such as the Gaussian process (GP) prior VAEs, can account for complex correlation structures across all data samples. While several methods have been proposed to learn standard VAEs from partially observed datasets, these methods fall short for conditional VAEs. In this work, we propose a method to learn conditional VAEs from datasets in which auxiliary covariates can contain missing values as well. The proposed method augments the conditional VAEs with a prior distribution for the missing covariates and estimates their posterior using amortised variational inference. At training time, our method accounts for the uncertainty associated with the missing covariates while simultaneously maximising the evidence lower bound. We develop computationally efficient methods to learn CVAEs and GP prior VAEs that are compatible with mini-batching. Our experiments on simulated datasets as well as on real-world biomedical datasets show that the proposed method outperforms previous methods in learning conditional VAEs from non-temporal, temporal, and longitudinal datasets.
Longitudinal Variational Autoencoder
(2021) Ramchandran, Siddharth; Tikhonov, Gleb; Kujanpaa, Kalle; Koskinen, Miika; Lahdesmaki, Harri
A4 Artikkeli konferenssijulkaisussa
Longitudinal datasets measured repeatedly over time from individual subjects, arise in many biomedical, psychological, social, and other studies. A common approach to analyse high-dimensional data that contains missing values is to learn a low-dimensional representation using variational autoencoders (VAEs). However, standard VAEs assume that the learnt representations are i.i.d., and fail to capture the correlations between the data samples. We propose the Longitudinal VAE (L-VAE), that uses a multi-output additive Gaussian process (GP) prior to extend the VAE's capability to learn structured low-dimensional representations imposed by auxiliary covariate information, and derive a new KL divergence upper bound for such GPs. Our approach can simultaneously accommodate both time-varying shared and random effects, produce structured low-dimensional representations, disentangle effects of individual covariates or their interactions, and achieve highly accurate predictive performance. We compare our model against previous methods on synthetic as well as clinical datasets, and demonstrate the state-of-theart performance in data imputation, reconstruction, and long-term prediction tasks.