Over-optimism in unsupervised microbiome analysis: Insights from network learning and clustering.

In recent years, unsupervised analysis of microbiome data, such as microbial network analysis and clustering, has increased in popularity. Many new statistical and computational methods have been proposed for these tasks. This multiplicity of analysis strategies poses a challenge for researchers, who are often unsure which method(s) to use and might be tempted to try different methods on their dataset to look for the "best" ones. However, if only the best results are selectively reported, this may cause over-optimism: the "best" method is overly fitted to the specific dataset, and the results might be non-replicable on validation data. Such effects will ultimately hinder research progress. Yet so far, these topics have been given little attention in the context of unsupervised microbiome analysis. In our illustrative study, we aim to quantify over-optimism effects in this context. We model the approach of a hypothetical microbiome researcher who undertakes four unsupervised research tasks: clustering of bacterial genera, hub detection in microbial networks, differential microbial network analysis, and clustering of samples. While these tasks are unsupervised, the researcher might still have certain expectations as to what constitutes interesting results. We translate these expectations into concrete evaluation criteria that the hypothetical researcher might want to optimize. We then randomly split an exemplary dataset from the American Gut Project into discovery and validation sets multiple times. For each research task, multiple method combinations (e.g., methods for data normalization, network generation, and/or clustering) are tried on the discovery data, and the combination that yields the best result according to the evaluation criterion is chosen. While the hypothetical researcher might only report this result, we also apply the "best" method combination to the validation dataset. The results are then compared between discovery and validation data. In all four research tasks, there are notable over-optimism effects; the results on the validation data set are worse compared to the discovery data, averaged over multiple random splits into discovery/validation data. Our study thus highlights the importance of validation and replication in microbiome analysis to obtain reliable results and demonstrates that the issue of over-optimism goes beyond the context of statistical testing and fishing for significance.

NetCoMi: network construction and comparison for microbiome data in R.

MOTIVATION: Estimating microbial association networks from high-throughput sequencing data is a common exploratory data analysis approach aiming at understanding the complex interplay of microbial communities in their natural habitat. Statistical network estimation workflows comprise several analysis steps, including methods for zero handling, data normalization and computing microbial associations. Since microbial interactions are likely to change between conditions, e.g. between healthy individuals and patients, identifying network differences between groups is often an integral secondary analysis step. Thus far, however, no unifying computational tool is available that facilitates the whole analysis workflow of constructing, analysing and comparing microbial association networks from high-throughput sequencing data. RESULTS: Here, we introduce NetCoMi (Network Construction and comparison for Microbiome data), an R package that integrates existing methods for each analysis step in a single reproducible computational workflow. The package offers functionality for constructing and analysing single microbial association networks as well as quantifying network differences. This enables insights into whether single taxa, groups of taxa or the overall network structure change between groups. NetCoMi also contains functionality for constructing differential networks, thus allowing to assess whether single pairs of taxa are differentially associated between two groups. Furthermore, NetCoMi facilitates the construction and analysis of dissimilarity networks of microbiome samples, enabling a high-level graphical summary of the heterogeneity of an entire microbiome sample collection. We illustrate NetCoMi's wide applicability using data sets from the GABRIELA study to compare microbial associations in settled dust from children's rooms between samples from two study centers (Ulm and Munich). AVAILABILITY: R scripts used for producing the examples shown in this manuscript are provided as supplementary data. The NetCoMi package, together with a tutorial, is available at https://github.com/stefpeschel/NetCoMi. CONTACT: Tel:+49 89 3187 43258; stefanie.peschel@mail.de. SUPPLEMENTARY INFORMATION: Supplementary data are available at Briefings in Bioinformatics online.

Maturation of the gut microbiome during the first year of life contributes to the protective farm effect on childhood asthma.

Growing up on a farm is associated with an asthma-protective effect, but the mechanisms underlying this effect are largely unknown. In the Protection against Allergy: Study in Rural Environments (PASTURE) birth cohort, we modeled maturation using 16S rRNA sequence data of the human gut microbiome in infants from 2 to 12 months of age. The estimated microbiome age (EMA) in 12-month-old infants was associated with previous farm exposure (ß = 0.27 (0.12-0.43), P = 0.001, n = 618) and reduced risk of asthma at school age (odds ratio (OR) = 0.72 (0.56-0.93), P = 0.011). EMA mediated the protective farm effect by 19%. In a nested case-control sample (n = 138), we found inverse associations of asthma with the measured level of fecal butyrate (OR = 0.28 (0.09-0.91), P = 0.034), bacterial taxa that predict butyrate production (OR = 0.38 (0.17-0.84), P = 0.017) and the relative abundance of the gene encoding butyryl-coenzyme A (CoA):acetate-CoA-transferase, a major enzyme in butyrate metabolism (OR = 0.43 (0.19-0.97), P = 0.042). The gut microbiome may contribute to asthma protection through metabolites, supporting the concept of a gut-lung axis in humans.

Composition of the cuticle of developing sweet cherry fruit.

The composition of wax and cutin from developing sweet cherry (Prunus avium) fruit was studied by GC-MS between 22 and 85 days after full bloom (DAFB). In this and our previous study, fruit mass and surface area increased in a sigmoidal pattern with time, but mass of the cuticular membrane (CM) per unit fruit surface area decreased. On a whole fruit basis, mass of CM increased up to 36 DAFB and remained constant thereafter. At maturity, triterpenes, alkanes and alcohols accounted for 75.6%, 19.1% and 1.2% of total wax, respectively. The most abundant constituents were the triterpenes ursolic (60.0%) and oleanolic acid (7.5%), the alkanes nonacosane (13.0%) and heptacosane (3.0%), and the secondary alcohol nonacosan-10-ol (1.1%). In developing fruit triterpenes per unit area decreased, but alkanes and alcohols remained essentially constant. The cutin fraction of mature fruit consisted of mostly C16 (69.5%) and, to a lower extent, C18 monomers (19.4%) comprising alkanoic, omega-hydroxyacids, alpha,omega-dicarboxylic and midchain hydroxylated acids. The most abundant constituents were 9(10),16-dihydroxy-hexadecanoic acid (53.6%) and 9,10,18-trihydroxy-octadecanoic acid (7.8%). Amounts of C16 and C18 monomers per unit area decreased in developing fruit, but remained approximately constant on a whole fruit basis. Within both classes of monomers, opposing changes occurred. Amounts of hexadecandioic, 16-hydroxy-hexadecanoic, 9(10)-hydroxy-hexadecane-1,16-dioic and 9,10-epoxy-octadecane-1,18-dioic acids increased, but 9,10,18-trihydroxy-octadecanoic and 9,10,18-trihydroxy-octadecenoic acids decreased. There were no qualitative and minor quantitative differences in wax and cutin composition between cultivars at maturity. Our data indicate that deposition of some constituents of wax and cutin ceased during early fruit development.

Changes in strain and deposition of cuticle in developing sweet cherry fruit.

Changes in surface area, deposition and elastic strain of the cuticular membrane (CM) were monitored during development of sweet cherry (Prunus avium L.) fruit. Fruit mass and surface area ('Sam') increased in a sigmoidal pattern between 16 and 85 days after full bloom (DAFB) with maximum rates of 0.35 g day(-1) and 0.62 cm(2) day(-1), respectively. Rates of total area strain, namely the sum of elastic plus plastic strain, were highest in cheek and stem cavity regions followed by stylar and suture regions. Rates of total uniaxial strain were higher in transverse, namely perpendicular to the stem/stylar axis, than in longitudinal direction, namely parallel to the stem/stylar axis. On a whole fruit basis CM mass remained essentially constant during fruit development. Mass of CM, dewaxed CM and wax per unit surface area decreased during development, particularly between 43 and 71 DAFB. There was no change in wax content of isolated CM. Up to 43 DAFB the surface area of isolated CM was similar to the area prior to excision indicating little elastic strain, but markedly decreased thereafter. Calculating elastic and plastic components of total strain of the CM revealed, that initial deformation up to 22 to 43 DAFB was mostly plastic. Thereafter, elastic strain was evident and both, elastic and plastic deformation, increased linearly with an increase in total strain. There was no consistent difference in the relative contribution of elastic strain to total strain between transverse and longitudinal directions, but both total and elastic strain were larger in the transverse direction. Abrading the CM had only little effect on fruit turgor. However, turgor decreased when the exocarp was cut indicating that the exocarp provided a significant structural shell of a mature sweet cherry fruit ('Regina'). Our data demonstrate, that (1) surface area expansion in sweet cherry fruit causes elastic and plastic strain of the CM, and (2) the onset of elastic strain coincided with the cessation of CM formation.

Studies on water transport through the sweet cherry fruit surface. 7. Fe3+ and Al3+ reduce conductance for water uptake.

The effects of the chloride salts LiCl, CaCl(2), MgCl(2), AlCl(3), EuCl(3), and FeCl(3) and the iron salts FeCl(2), FeCl(3), Fe(NO(3))(3), FeSO(4), and Fe(2)(SO(4))(3) on water conductance of exocarp segments (ES) and rates of water uptake into detached sweet cherry fruit (Prunus avium L. cv. Adriana, Early Rivers, Namare, Namosa, and Sam) were studied. ES were excised from the cheek of mature fruit and mounted in stainless steel diffusion cell; water penetration was monitored gravimetrically from donor solutions containing the above mineral salts into a PEG 6000 (osmolality = 1.14 osM, pH 4.8, 25 degrees C) receiver solution. Conductance of ES was calculated from the amount of water taken up per unit of surface area and time by dividing by the gradient in water activity across ES. LiCl, CaCl(2), MgCl(2), FeCl(2), and FeSO(4) had no significant effect on conductance, but AlCl(3), FeCl(3), Fe(NO(3))(3), and Fe(2)(SO(4))(3) significantly reduced conductance compared to water only as a donor. Also, EuCl(3) lowered conductance; however, this effect was not always significant. Effects of salts on water conductance of ES and rates of water uptake into detached fruit were closely related (R 2 = 0.97***). Upon application of an FeCl(3)-containing donor conductance decreased instantaneously. FeCl(3) concentrations of <6.6 x 10(-)(4) M had no effect on conductance, but concentrations at or above this threshold decreased conductance. FeCl(3) lowered water conductance at a receiver pH of 4.8, but not at pH < or =2.6. The effect of FeCl(3) on conductance was largest in cv. Namare and smallest in cv. Adriana. There was no significant effect of FeCl(3) on conductance for transpiration. Formation of aluminum and iron oxides and hydroxides in the exocarp as a result of a pH gradient between donor and receiver solution is discussed as the potential mechanism for Fe(3+) and Al(3+) reducing conductance for water uptake.

Studies on water transport through the sweet cherry fruit surface: III. Conductance of the cuticle in relation to fruit size.

Rain-cracking of sweet cherry fruit has been related to water absorption through the fruit surface and large fruit has been reported to be more susceptible to cracking than small fruit. Therefore, the effect of fruit size on water conductance of the cuticular membrane (CM) of exocarp segments excised from cheek, suture or stylar end region of mature sweet cherry fruit (Prunus avium L. cv. Sam) was investigated. Segments consisting of epidermis, hypodermis and several layers of mesocarp cells were mounted in diffusion cells filled with deionized water. Mass loss due to transpiration was monitored gravimetrically during an 8-h incubation period (25 +/- 2 degrees C) over dry silica in the dark. Conductance was calculated from the amount of water transpired per unit surface area and time divided by the difference in water vapour concentration across the segment. For an average size cv. Sam sweet cherry CM conductance was 1.06 x 10-4, 0.91 x 10-4 and 2.09 x 10-4 m s-1 in cheek, suture and stylar end region, respectively. Fruit size had no significant effect on conductance in cheek or suture regions, but for the stylar end region conductance was positively related to fruit size. Stomatal density in the cheek, but not the suture or stylar end region increased as fruit size increased. The area of the stylar scar was positively related to fruit size. Conductance of the stylar scar averaged 37.6 +/- 4.0 x 10-4 m s-1 and was 54-fold higher than that of the CM between stomata in the cheek region (mean 0.69 x 10-4 m s-1). Conductance calculated on a whole fruit basis is estimated to increase by 108% as fruit size increases from 6 to 12 g. Increased conductance on a whole fruit basis may be attributed to increased fruit surface area and increased conductance per unit fruit surface area, particularly in the stylar end region.