Based on this linear model, we developed a all markers’ and marker pairs’ effects, those with higher −log10 PATOWAS pipeline to analyze traits through multiple ome- values indicate markers or marker pairs that are more relevant wide association studies. Therefore, the proposed model and to the phenotypic trait. PATOWAS can be used to study not only GWAS for G2P but In the present study, we sequentially submitted three omic also TWAS for T2P and MWAS for M2P, which is progress marker datasets to PATOWAS to analyze the two field traits, toward an integrative omics . YIELD and KGW. We downloaded the results after completion of To test this presumption and verify our consideration, we the analyses. Based on these results, multiple associative omics used PATOWAS to analyze the rice RIL datasets with two and the biological insight can be compared and integrated. For agronomic traits and three different omics markers. PATOWAS example, the combination of 1D association mapping across accepts 2D omics marker matrix data and 1D phenotypic trait G2P and T2P can help identify the genotype and expressed gene data as inputs .
PATOWAS results for one specific transcript markers with consistent physical positions; comparison associative omics mainly include three parts: variance component of the metabolites from 1D M2P association mapping can analysis for the partition of phenotypic variance, a 1D association uncover the biochemical relevance of tissue-specific metabolites map for the direct biological markers,30 in pot and a 2D association map and traits to be analyzed; and the investigation of major for the interaction of biological marker pairs . Of the biomarker pairs from 2D association mapping can be used to three variance components, the additive component for the build an association network. All these together provide a systems markers’ direct effects and the additive–additive component for biology view into the analyzed traits, leading toward an answer the marker pairs’ interaction effects are biologically meaningful about how genes, transcripts, proteins, and metabolites work and can be explained by the linear model. The higher the sum of together to produce an observable phenotype.Based on the variance component analysis results, we generated six pie charts displaying the three variance components of the two traits across associative genomics, associative transcriptomics, and associative metabolomics . We found that the two biologically meaningful variance components accounted for nearly all of the phenotypic trait variance in associative transcriptomics and associative metabo lomics but not in associative genomics. Also, YIELD was a more complex trait than KGW, as the two biologically meaningful variance components accounted for only 66% of the total phenotypic variance in associative genomics but nearly 100% of the total phenotypic variance in associative transcriptomics and metabolomics .
These findings demonstrate that a chain of environmentally responsive genes and metabolites can be observed and explained at the transcriptomic and metabolomic levels but not at the genomic level. Here we noticed that the marker number for transcripts was obviously one-order of scale higher than the other two. Consider the marker-by-marker interactions: The pairwise number of transcripts will reach to ~250 million, which is about two-order of scale larger than the other two kinds of omic markers. To test whether the higher ratio of biological explanatory components observed in the TWAS result is not due to the larger numbers of transcripts used in TWAS, we further produced a reduced transcript gene set with a number scale comparable to the genotypes and metabolites. We separately submitted the reduced transcript gene set to PATOWAS and checked the variance component analysis result. The procedures to generate a reduced gene set are described as follows: First we mapped the 22,584 transcript genes into the 1619 genotype bins ; one genotype bin may contain none to hundreds of transcript genes. Based on the 1D association mapping result, at most only one representative transcript in one bin was selected. We chose the transcript with the highest −log10 as the representative transcript of a genotype bin. Then we generated a reduced transcript gene set for each phenotypic trait, which essentially is a data matrix with a dimension of 1543 × 210 . Its number of markers was comparable to those in the analyzed genotypes and metabolites. The same approaches were also used to generate two positional comparable 1D G2P and T2P association mapping results in the following section.
We submitted the reduced transcript data and the two phenotypic traits, KGW and YIELD, to PATOWAS for further study. Based on the variance component analysis results, two additional pie charts displaying the three variance components of the two traits in associative transcriptomics were plotted . Again, we observed that the two biologically meaningful components explained nearly 100% of the phenotypic variance, with only a fluctuation between the two components. Thus, we conclude that the much larger numbers of transcripts used in TWAS is not the reason for the higher explanatory ratio of phenotypic variance in associative transcriptomics.Modern GWAS application often involves a panel with hundreds of thousands, or even millions, of genetic variants under only several hundred individual samples. The statistical modeling of such cases is usually challenging because the sample size is substantially smaller than the number of covariates. This is well-known as a “large p small n” problem and requires careful assessment of the statistical characteristics. Our proposed method really can explain more of phenotypic variance, but the cost is that it generates a large number of pairwise covariates. Therefore, it is worthwhile to assess the heritability of the proposed LMM, particularly at the high dimensional data. First, the predictability that is represented by the squared correlation coefficient between the observed and predicted phenotypic value was applied. The squared correlation is approximately equal to R2 = 1−PRESS/SS, where PRESS is the predicted residual error sum of squares and SS is the total sum of squares of the phenotypic values. In principle, we treated each transcript or metabolite marker as an intermediate phenotypic trait and predicated all of these intermediated phenotypic values from all the genotypic data. Therefore, each transcript or metabolite will have an R2 value, predictability . We then used the HAT method to calculate the PREDs for all transcripts and metabolites , applied a series of variable thresholds to the PREDs,dutch bucket for tomatoes and selected the transcript and metabolite markers. Finally, we submitted the subsets of selected transcript genes and metabolites to PATO WAS for variance component analysis and calculated the broad sense heritability, H. Figure 3 shows the assessment result of the broad-sense heritability with the selected markers by PRED thresholding. We found that the number of selected markers continued decreasing as the PRED threshold increased; however, the broad-sense H provides us with a very different perspective of different traits and different associative omics. It needs only ~1000 and fewer than 100 transcripts to explain more than 97% of the phenotypic variance in traits YIELD and KGW, respectively. In associative metabolomics, only 30 metabolites are enough to explain more than 90% of the phenotypic variance. In general, trait KGW is more conserved than trait YIELD, and associative metabolomics is more conserved than associative transcriptomics. Variance component analysis provides us with a big picture by partitioning the phenotypic variation into three components. The two biologically meaningful components for individual markers’ direct effects and the marker pairs’ interaction effects can be further illustrated by 1D and 2D association mapping, respectively.For trait YIELD, three 2D association mapping results were analyzed, and each association matrix was illustrated as a scaled image with pseudocolor . By com parison, we found that genotypic markers were neighbor dependent, as evidenced by the clustering of dots, whereas expressed transcript gene and metabolite markers were neighbor independent, as evidenced by a random distribution of dots. This phenomenon could be explained by the existence of linkage disequilibrium blocks in population genetics. We are usually interested in the significant ≥ Significance_Th) marker pairs instead of all the marker pairs. Similar to 1D association mapping, we could set a significance threshold to generate a binarized version of the 2D association matrix.
We further zoomed in to a specified local region for each associative omics and found that associative genomics demonstrated a 2D local rectangular array while the associative transcriptomics and associative metabolo mics showed a 1D local strip . The specificity of the 2D local structure pattern for associative genomics was due to the existence of LD blocks in genomics level. Further, the dimension size of 2D local rectangular array corresponds to the LD block size.In 2015 the global average concentration of carbon dioxide in the atmosphere reached a record high of 400 μmol CO2 mol-1 air and if current trends continue by the end of this century it could even surpass 800 μmol CO2 mol-1 air. Rising [CO2] is largely responsible for changes in our climate including increased temperatures and altered precipitation patterns. These changes in weather patterns will ultimately influence crop productivity and are predicted to be particularly detrimental to summer crops, such as maize, which will likely experience severe episodes of drought. Maize represents an essential part of the world’s grain food and feed supply, and the majority of the maize cropping systems depends on natural precipitation. Maize uses the C4 photosynthetic mechanism which is not limited by [CO2], and therefore, yields will only benefit from rising [CO2] under conditions of drought when the indirect effect of reduced stomatal conductance enhances the plants water-use efficiency allowing photosynthesis to continue despite limited water conditions. Nevertheless, in addition to abiotic stress, plant diseases and insect pests are also major limiting factors of maize productiv ity, yield, and quality; however, our understanding of how the combination of both elevated [CO2] and drought will affect maize susceptibility to biotic stressors is limited. The mycotoxigenic fungal pathogen, Fusarium verticillioides not only reduces the maize yield by causing rot in all parts of the plant but also produces carcinogenic polyke tide-derived mycotoxins termed fumonisins which render harvested grain unsafe for human or animal consumption. Mycotoxins, such as fumonisins, are among the top food safety concerns with regard to climate change because environmental conditions predicted for the future are important factors that contribute to fumonisin contamination. Warmer temperatures increase evapotranspiration further intensifying drought which has been shown to correlate with Fv disease development and enhance fumonisin accumulation in grain. Recently, we demonstrated that elevated [CO2] also enhances maize susceptibility to Fv infection, but the increase in fungal biomass did not correspond with greater fumonisin levels resulting in an overall reduction in fumonisin per unit fungal biomass. Following Fv inoculation, the accumulation of maize soluble sugars, free fatty acids, lipox ygenase transcripts, jasmonic acid and salicylic acid phytohormones, and ter penoid phytoalexins was dampened at elevated [CO2]. An influx of fatty acid substrate is essential for the burst of JA that initiates the defense signaling process. JA and other oxyli pins are synthesized from free fatty acids through the LOX pathway. The fatty acids are oxidized by LOX enzymes at either the 9 or 13 carbon position to produce 9-LOX or 13-LOX oxylipins, respectively.The defense related functions of 9-LOX metabolites are not well characterized, but they have been implicated in the stimulation of mycotoxin production. Elevated [CO2] appears to effect both 9- and 13-LOX oxylipin biosynthesis at the level of fatty acid sub strate supply and LOX-gene transcription. Lower concentrations of defensive phytochemicals, such as the zealexin and kauralexin terpenoid phytoalexins, due to compromised JA biosynthesis and signaling is consistent with increased Fv proliferation. Additionally, a damp ened response of 9-LOX metabolites could reduce the ratio of fumonisin per Fv biomass. Elevated [CO2] similarly enhances C3 crop susceptibility to herbivory by compromising LOX-gene transcription, JA biosynthesis and JA-dependent antiherbivore defenses. However, in C3 crops JA regulated defenses appeared to be compromised by an antagonistic boost in SA production, which does not occur in maize. Furthermore, the effects of elevated [CO2] were negated when soybean plants were simultaneously exposed to drought stress. Whether drought will negate the effects of elevated [CO2] on maize susceptibility to Fv is unknown, and it is unclear what the interactive effects of elevated [CO2] and drought will do to fumonisin levels. Although individual stress responses display measurable specificity, plants are frequently simultaneously challenged by several stress factors resulting in the activation of multiple signals that engage in cross-talk and alter individual responses.