Limits...
Genome-wide association mapping of growth dynamics detects time-specific and general quantitative trait loci.

Bac-Molenaar JA, Vreugdenhil D, Granier C, Keurentjes JJ - J. Exp. Bot. (2015)

Bottom Line: Genome-wide association (GWA) mapping of the temporal growth data resulted in the detection of time-specific quantitative trait loci (QTLs), whereas mapping of model parameters resulted in another set of QTLs related to the whole growth curve.The positive correlation between projected leaf area (PLA) at different time points during the course of the experiment suggested the existence of general growth factors with a function in multiple developmental stages or with prolonged downstream effects.In addition, the detection of QTLs without obvious candidate genes implies the annotation of novel functions for underlying genes.

View Article: PubMed Central - PubMed

Affiliation: Laboratory of Plant Physiology, Wageningen University, Droevendaalsesteeg 1, 6708 PB Wageningen, The Netherlands Laboratory of Genetics, Wageningen University, Droevendaalsesteeg 1, 6708 PB Wageningen, The Netherlands.

No MeSH data available.


Comparison of the goodness-of-fit for the three growth models used: exponential function with one (Expo1) or two (Expo2) parameters, and Gompertz function (Gom). Plot of the measured PLA on days 8, 11, 14, 16, 18, 20, 22, 24, 25, 26, 27, and 28 against the predicted PLA on the same days. The black line represents y=x (PLA measured=PLA predicted).
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC4585414&req=5

Figure 4: Comparison of the goodness-of-fit for the three growth models used: exponential function with one (Expo1) or two (Expo2) parameters, and Gompertz function (Gom). Plot of the measured PLA on days 8, 11, 14, 16, 18, 20, 22, 24, 25, 26, 27, and 28 against the predicted PLA on the same days. The black line represents y=x (PLA measured=PLA predicted).

Mentions: For each model, the goodness-of-fit was evaluated (Fig. 4; Supplementary Table S2 at JXB online). As expected, r2 was on average higher when more parameters were introduced into the model (Supplementary Table S2). Expo1 predictions were in general too low at small PLA and too high at large PLA (Fig. 4), which indicates that this model is too simplistic. Interestingly, the differences in goodness-of-fit between Expo2 and Gom were not large, emphasizing that most plants in this experiment do not reach ‘K’ and that the growth rate thus does not decrease significantly during the duration of the experiment. So, for most plants in this experiment, determinate growth cannot be concluded from the PLA data collected. This was supported by the smaller confidence intervals for the parameters of Expo2 compared with the parameters of Gom (Supplementary Table S2). For Amax in particular, very large confidence intervals were observed. Based on Fig. 4, the confidence intervals and the principle of parsimony, stating that the simplest of two competing models is to be preferred, Expo2 was chosen to be used in the GWA analyses. This model is counter-intuitive because it describes indeterminate growth, while it is known that the Arabidopsis rosette follows determinate growth (Leister et al., 1999). In this case, a model describing determinate growth, such as the Gompertz function, results in parameters that are more informative (or speculative) for growth outside than inside the experimental window. Growth models that describe an S-curve always contain a parameter representing the final rosette size, and the other parameters that are estimated are dependent on that parameter. In the present case, Gom, which describes an S-curve, would have functioned as a (not very reliable) predictive model instead of a descriptive model as was aimed for. If curve fitting using the growth model results in reliable fits, as it did for most plants in this experiment, it allows for comparison of plants which differ in developmental timing, growth rate, and plant size. However, this comparison only leads to valuable insight if the right model is chosen. Conclusions based on a non-optimal model should be interpreted carefully as they can easily become very speculative (Tessmer et al., 2013).


Genome-wide association mapping of growth dynamics detects time-specific and general quantitative trait loci.

Bac-Molenaar JA, Vreugdenhil D, Granier C, Keurentjes JJ - J. Exp. Bot. (2015)

Comparison of the goodness-of-fit for the three growth models used: exponential function with one (Expo1) or two (Expo2) parameters, and Gompertz function (Gom). Plot of the measured PLA on days 8, 11, 14, 16, 18, 20, 22, 24, 25, 26, 27, and 28 against the predicted PLA on the same days. The black line represents y=x (PLA measured=PLA predicted).
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC4585414&req=5

Figure 4: Comparison of the goodness-of-fit for the three growth models used: exponential function with one (Expo1) or two (Expo2) parameters, and Gompertz function (Gom). Plot of the measured PLA on days 8, 11, 14, 16, 18, 20, 22, 24, 25, 26, 27, and 28 against the predicted PLA on the same days. The black line represents y=x (PLA measured=PLA predicted).
Mentions: For each model, the goodness-of-fit was evaluated (Fig. 4; Supplementary Table S2 at JXB online). As expected, r2 was on average higher when more parameters were introduced into the model (Supplementary Table S2). Expo1 predictions were in general too low at small PLA and too high at large PLA (Fig. 4), which indicates that this model is too simplistic. Interestingly, the differences in goodness-of-fit between Expo2 and Gom were not large, emphasizing that most plants in this experiment do not reach ‘K’ and that the growth rate thus does not decrease significantly during the duration of the experiment. So, for most plants in this experiment, determinate growth cannot be concluded from the PLA data collected. This was supported by the smaller confidence intervals for the parameters of Expo2 compared with the parameters of Gom (Supplementary Table S2). For Amax in particular, very large confidence intervals were observed. Based on Fig. 4, the confidence intervals and the principle of parsimony, stating that the simplest of two competing models is to be preferred, Expo2 was chosen to be used in the GWA analyses. This model is counter-intuitive because it describes indeterminate growth, while it is known that the Arabidopsis rosette follows determinate growth (Leister et al., 1999). In this case, a model describing determinate growth, such as the Gompertz function, results in parameters that are more informative (or speculative) for growth outside than inside the experimental window. Growth models that describe an S-curve always contain a parameter representing the final rosette size, and the other parameters that are estimated are dependent on that parameter. In the present case, Gom, which describes an S-curve, would have functioned as a (not very reliable) predictive model instead of a descriptive model as was aimed for. If curve fitting using the growth model results in reliable fits, as it did for most plants in this experiment, it allows for comparison of plants which differ in developmental timing, growth rate, and plant size. However, this comparison only leads to valuable insight if the right model is chosen. Conclusions based on a non-optimal model should be interpreted carefully as they can easily become very speculative (Tessmer et al., 2013).

Bottom Line: Genome-wide association (GWA) mapping of the temporal growth data resulted in the detection of time-specific quantitative trait loci (QTLs), whereas mapping of model parameters resulted in another set of QTLs related to the whole growth curve.The positive correlation between projected leaf area (PLA) at different time points during the course of the experiment suggested the existence of general growth factors with a function in multiple developmental stages or with prolonged downstream effects.In addition, the detection of QTLs without obvious candidate genes implies the annotation of novel functions for underlying genes.

View Article: PubMed Central - PubMed

Affiliation: Laboratory of Plant Physiology, Wageningen University, Droevendaalsesteeg 1, 6708 PB Wageningen, The Netherlands Laboratory of Genetics, Wageningen University, Droevendaalsesteeg 1, 6708 PB Wageningen, The Netherlands.

No MeSH data available.