We then performed a detailed comparison of several gene properties between modules, allowing for a less biased and more powerful analysis.
Notably, our analysis corroborated the hourglass pattern at the regulatory level, with sequences of regulatory regions being most conserved for genes expressed in mid-development but not at the level of gene sequence, age, or expression, in contrast to some previous studies.
The early conservation model was supported with gene duplication and birth that were the most rare for genes expressed in early development. Finally, for all gene properties, we observed the least conservation for genes expressed in late development or adult, consistent with both models. Overall, with the modular approach, we showed that different levels of molecular evolution follow different patterns of developmental constraints. Thus both models are valid, but with respect to different genomic features.
This gave rise to the hourglass model, which predicts the highest developmental constraints during mid-embryogenesis. In the last decade, a large effort has been made to uncover the relation between developmental constraints and the evolution of genome. Several studies reported gene characteristics that change according to the hourglass model, e.
Here, we first show that some of the previous conclusions do not hold out under detailed analysis of the data.
Then, we discuss the disadvantages of the standard evo-devo approach, i. Results of such analysis are biased by genes expressed constantly during development housekeeping genes.
To overcome this limitation, we use a modularization approach, which reduces the complexity of the data and assures independency between the sets of genes which are compared. We identified distinct sets of genes modules with time-specific expression in zebrafish development and analyzed their conservation of sequence, gene expression, and regulatory elements, as well as their age and orthology relationships.
Interestingly, we found different patterns of developmental constraints for different gene properties. Only conserved regulatory regions follow an hourglass pattern. PLoS Genet 9 4 : e Editor: Gregory S. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist. Developmental constraints have been suggested to play an important role in shaping the evolution of embryonic development in animals. Briefly, the concept of developmental constraints assumes that the scope of developmental mechanisms limits the set of phenotypes that may evolve.
Thus, morphological similarities between embryos of different species could reflect these underlying constraints [1]. Two main models of embryonic developmental constraints have been put forward. The early conservation model predicts that the highest developmental constraints occur at the beginning of embryogenesis.
This corresponds to von Baer's third law [2] , postulating that embryos of different species progressively diverge from one another during ontogeny. However, in modern times, the highest morphological similarity between embryos of different species was observed in the phylotypic stage i. Consequently, Duboule [6] and Raff [7] proposed the so-called hourglass model, which has since become widely accepted see, e.
It predicts the highest developmental constraints during mid-embryogenesis. At the genomic level, the hourglass model was originally linked to the expression of Hox genes in animals [6]. More recently, the emphasis has shifted to the relation, if any, between developmental constraints and the evolution and function of the genome reviewed in [9].
Different studies have reported several characteristics supporting the hourglass model in animals on the genomic level. Hazkani-Covo et al. Similarly, Irie and Kuratani [13] reported the highest expression conservation between zebrafish, frog, chicken and mouse, for genes expressed in mid-development.
Very recently, the hourglass model was argued to hold also for plants embryogenesis with respect to gene age and sequence conservation [14]. However, some of these results do not hold out under detailed analyses see Box 1 and Text S1. For example, applying a standard log-transformation [15] , [16] to microarray signal intensities used in [11] changes the reported pattern such that it no longer supports the hourglass model Figure 1.
Moreover, other studies have also found genetic patterns supporting an early conservation model [17] , [18]. A higher TAI value implies that evolutionary younger genes are preferentially expressed at the corresponding time-point.
The pink shaded area indicates the phylotypic stage. Colors of the curves reflect the main developmental periods and correspond to the colors used in [11]. Note that the microarray signal intensity values used in [11] displayed a log-normal distribution and spanned from 1 to Figure S1. Using these values to calculate TAI made the weights of phylogenetic ranks differ by five orders of magnitude between lowly and highly expressed genes.
Consequently, only the most expressed genes Figure S2 , and potentially outliers Figure S3 , contributed to the hourglass pattern discovered with TAI.
We found that applying a standard log-transformation to the intensity values changes the pattern, which then indicates older genes being expressed preferentially in early development Figure 1. The use of log-transformed data for microarray intensities is generally encouraged [15] , [16] because it keeps the biological signal, while removing dependency between variance and intensity of the analyzed signals.
We also discuss in Text S1 the study of Quint et al. In most of the studies of developmental constraints the authors compared descriptive statistics of all genes across all developmental time-points e.
Such an approach introduces dependencies between the sets of genes which are compared, and consequently can produce results biased by genes expressed at many time-points.
For example, housekeeping genes contribute to the average gene expression at all time points, and hence dilute trends. To overcome this essential problem, we have used a modularization approach, which we applied to the recently published transcriptome data of zebrafish development [11]. We decomposed the genes into independent sets, i. This decomposition allowed us to compare only sets of genes that have specific functions during embryonic development.
For each of the seven modules, we studied five properties of its genes: 1 gene sequence conservation, 2 gene age, 3 gene expression conservation, 4 gene orthology relationships, and 5 regulatory elements conservation. Here, we show that different levels of molecular evolution follow different patterns of developmental constraints. First, the regulatory elements are most conserved for transcription factors expressed at mid-development, consistent with the hourglass model.
Contrary to what has been reported previously [10] , [11] , [13] , we did not detect the hourglass pattern for gene sequence, age and expression. Second, constraints on gene duplication and on new gene introduction are the strongest in early development, supporting the early conservation model consistent with [17].
Finally, all gene properties displayed the least conservation in late development and adult, which is in agreement with both models of developmental constraints. Our goal was to analyze the developmental constraints acting on different gene properties. To this end we identified and analyzed groups of genes co-expressed during distinct developmental stages. The ISA is a modularization algorithm that finds genes with similar expression profiles and groups them into so-called transcription modules.
In order to detect modules of genes with specific expression during the zebrafish development, we initialized the ISA with seven idealized expression profiles that corresponded to successive developmental stages see Text S1 and Figure S Overall, the modules covered the entire development.
The phylotypic stage in which the hourglass model predicts the highest evolutionary constraint corresponds to the segmentation and pharyngula modules. We will refer to these two modules as phylotypic modules. A Zebrafish ontogeny drawings of the embryos are based upon sketches and photographs from [49].
B Median, 25th and 75th percentiles of expression value of genes in modules. Red bars denote the condition scores assigned to developmental points by the ISA. The adjacent modules partially overlapped in their gene content. In order to allow for unbiased cross-module comparisons, genes belonging to two modules were kept in the one with the highest ISA gene score see Methods ; this concerned genes in total.
The seven modules, i. Overall, different genes were present in these modules, which implies a significant reduction of the number of genes being analyzed in comparison to the original data genes on the microarray.
In particular, the ISA removed the bias related to the genes expressed uniformly across development like housekeeping genes. We verified the function of genes in modules detected by the ISA by comparing them to relevant known lists of genes. We confirmed the relevance of the segmentation and pharyngula modules by verifying that they were enriched in Hox genes 24 and 7 genes vs.
We did not have any gold standard for genes expressed at the late stages of development. However, since the early and phylotypic modules were enriched in genes with relevant functions, we are confident that the same is true for the late modules. Moreover, gene ontology GO enrichment analysis confirmed that genes from the modules were enriched in functions relevant to the respective developmental stages. The pharyngula module was enriched in genes associated with cell differentiation, and anatomical structure development.
Finally, the adult module was enriched in genes involved in responses to environment, although not significantly Table S2. We checked whether the sequences of genes from different modules evolved under different selective pressure. To this end, we calculated the non-synonymous to synonymous substitution ratios for genes in the modules and asked if the ratio was significantly lower for any of them.
With the early conservation model, we would expect the lowest values for genes from early modules. Whereas with the hourglass model, we would expect the lowest values for genes from the phylotypic modules.
In the other four modules covering embryonic development the median was lower than the median for all genes Figure 3A , and the difference was significant for all but the segmentation module randomization test, for the gastrula, pharyngula and larva modules. In the juvenile module, the median was significantly higher than the median for all genes randomization test,. In the adult module, the median was also higher than the median for all genes, but the difference was not significant.
When analyzing separately sites under purifying selection or evolving neutrally, we also find weaker purifying selection during post-embryonic stages see Text S1 and Figure S A Box and Whisker plot showing non-synonymous to synonymous substitution ratios for genes in the modules. The dotted line denotes median for all genes. The dash-dotted lines denote confidence interval for the median. B Observed minus expected frequencies of age of genes in modules. C Observed minus expected frequencies of orthology type between zebrafish and mouse for genes in modules.
D Mean expression level of zebrafish genes in modules, and their one-to-one orthologs in mouse in six developmental metastages.
The transition between the two mouse data sets is denoted with the vertical dashed line. The Pearson's correlation coefficients for zebrafish and mouse expression profiles are reported for every module. The dash-dotted lines denote confidence interval for the expected number of TFs in modules. These results were consistent with the study by Roux and Robinson-Rechavi [17] , who also reported equally low values during the entire zebrafish embryogenesis, and a small increase in mid-larva, juvenile and adult.
In contrast, Hazkani-Covo et al. However, the trend was not significant. In [17] some evidence for early conservation was reported in mouse. Projecting the genes from zebrafish modules to mouse-human orthologs, we found equal conservation across development Figure S Overall, data analyses support similar evolutionary constraints on sequences of genes expressed during whole embryogenesis of zebrafish, while for mouse more developmental data is needed to be conclusive.
The differences in age of genes expressed during different stages of the development have been suggested to be a good indicator of evolutionary constraints [11] , [25]. Thus, we investigated the age of genes belonging to different modules. Next, for each module we calculated the age distribution of its genes, i. The younger Bilateria genes were overrepresented in the phylotypic modules The youngest genes were overrepresented in the late modules e. Yet, that result does not hold for log-transformed gene expression levels Box 1 , and is not recovered with measures of gene age other than the transcriptome age index see Text S1 and Figure S6.
With the modular approach we observed that the age of expressed genes decreased throughout ontogeny. This pattern suggests that the oldest evolutionary stages tend to express the oldest genes. Both gene duplication and gene loss can impact phenotypic evolution [26] — [30].
The outcome of these events can be summarized by the resulting gene family size. Consequently, constrained developmental stages should display less changes in gene family size than other stages. To test this hypothesis, for each zebrafish module we calculated the number of its genes that were in 1 one-to-one, 2 one-to-many, 3 many-to-many, and 4 no orthology relation to mouse genes i. We compared the observed distributions with the distribution of the ortholog relationships for all genes.
Moreover, the pattern of variation itself differed across different modules. The number of one-to-one orthologs decreased throughout development Figure 3C. In contrast, the number of genes with no orthologous relationship increased throughout development Figure 3C. It was significantly higher than expected only in the juvenile and adult modules A similar pattern was observed for many-to-many orthologs Finally, the number of one-to-many orthologs was higher than expected only in the larva module These results were consistent with [17] in which the genes retained in duplicates after the teleost-specific whole genome duplication were reported to have low expression early in the development.
Here, we recovered an analogous pattern with the modular approach, showing that the genes expressed early in the development are retained in duplicates less often than genes expressed later.
Note that our observation is not limited to whole genome duplication. In addition, we detected the highest number of novel genes amongst genes expressed late in the development. Changes in gene expression are one of the main sources of morphological variation [32] — [34]. The developmental constraints on gene expression might differ from those on the gene sequence [35] — [37].
Thus, for each module, we compared the mean expression profile of its genes with the mean expression profile of their one-to-one orthologs in mouse. We used two different data sets [13] , [38] with expression values of mouse genes during the development. The use of two data sets was necessary, because there does not exist a single experiment covering the entire mouse development.
The incompatibility of the two microarrays impaired the statistical strength of the analysis. For this reasons the results reported here should be regarded rather as qualitative than quantitative.
Since homology cannot be defined for individual developmental stages between zebrafish and mouse, we first mapped every time point to its broad metastage defined in Bgee database [39] Figure 4. Next, we calculated the mean expression level in every metastage.
This resulted in six expression values for each gene during the development of mouse and zebrafish: zygote, cleavage, blastula, neurula, organogenesis, and post-embryonic stage. Note that the mouse microarrays did not cover the gastrula stage at all. For each module we calculated the Pearson's correlation between the mean expression of its genes and their mouse orthologs across the six metastages. Nevertheless, there exists a plausible, biological interpretation of the differences in gene expression between the early stages of zebrafish and mouse development.
Zebrafish and mouse form two different embryological structures during blastulation, a blastula and a blastocyst, respectively. The blastocyst is a mammalian innovation that consists of an embryoblast that develop into structures of the fetus and a trophoblast that form the extraembryonic tissue. In contrast, there is no extraembryonic tissue in zebrafish. Overall, the lack of correlation between gene expression for the early stages of mouse and zebrafish development could be explained by these structural differences.
The hourglass pattern has been found outside of animals, in fungi Cheng et al. The phylogenetic breadth of the organisms where this pattern has been found has led to the suggestion that this is a fundamental characteristic of the ontology of multicellular organisms Cheng et al.
In addition, the pattern has recently been reported in biological processes other than developmental ontology, such as organogenesis Drost et al. The apparent ubiquity of the hourglass pattern has lead to the hypothesis that this pattern is not just a characteristic of developmental ontology, but a general pattern of evolution for complex biological processes Drost et al.
To date, only a scattered sampling of organisms in the tree of life have been studied using this approach. For example, studies of the developmental phylotranscriptomics in plant and fungi both have only been reported from one species Quint et al. Similarly, most animal studies focus on Deuterostomia and Ecdysozoa, two clades of Bilateria, whereas Spiralia, the other major bilaterian group, has rarely been examined.
Spiralia is an ancient and highly diverse clade of protostome animal groups. The adult body plans of spiralians are extremely diverse, but there are two developmental stages that might arguably be considered phylotypic stages.
The first is the spiral cleavage program in early development, which inspired the name Spiralia when it was recognized as homologous between molluscs, annelids and flatworms Schleip This cleavage pattern is characterized by highly regular asymmetries and cleavage angles during early divisions fig. It is also associated with strong similarities in the fate map of the blastula produced by these divisions.
For instance, precisely the same cell in the lineage generates endomesoderm in all groups with spiral cleavage reviewed in Lambert Since Schleip, the conserved spiral cleavage has also been recognized in nemerteans, and modified spiral cleavage has possibly been detected in other groups of spiralians reviewed in Hejnol ; Lambert ; Vellutini et al. This level of conservation within and between phylum-level groups in cleavage pattern and early cell fate specification is unparalleled across the animal kingdom.
The Spiralia. A The clade Spiralia within the metazoan phylogeny based on Laumer et al. Most previous studies examined taxa in the Deuterostomia and Ecdysozoa clades.
B Spiral cleavage shown in an 8 cell and 16 cell embryo, showing daughter cell asymmetry and the alternating angles of division that characterize this mode of development. The animal pole is up. Homologous spiral cleavage is recognized in molluscs, annelids, nemerteans, and platyhelminth flatworms.
C Generalized morphology of trochophore larva, a free-swimming planktonic larva with bands of cilia for locomotion and feeding. Homologous trochophore larvae are recognized in the molluscs and annelids, and this has been proposed to be the phylotypic stage in these groups see text. The other candidate for a phylotypic stage is the distinctive trochophore larva, found in molluscs and annelids fig. This pelagic larva has a pre-oral circumferential ciliary band that is used for feeding and swimming, an apical tuft of cilia at the anterior end, and a posterior ring of cilia called the telotroch.
The trochophore larva stage generally occurs while organogenesis is ongoing, and the basic bodyplan is emerging. The trochophore stage has in fact been proposed to be the phylotypic stage for molluscs and annelids Cohen and Massey ; Slack ; Shigeno et al.
Compared with the other two clades of bilaterian animals, there are fewer spiralian model systems and less genomic data available for these animals.
One recent paper recovered the hourglass pattern from spiralian transcriptomic data sets Xu et al. Here, we revisit this question using a refined calculation method and recently accumulated spiralian genomic and transcriptomic data, and find that spiralian development does not follow the predictions of the hourglass model, but instead shows a striking reverse hourglass pattern. Our results show that in spiralians, one of the previously proposed phylotypic stages—the trochophore, is the least conserved in terms of the evolution of expressed genes.
This work highlights the uniqueness of spiralian development and adds important refinements to phylotranscriptomics approach. To explore the relative ages of genes expressed at different developmental stages in a spiralian, we started by calculating the TAI of a bivalve mollusc, the pacific oyster C. This species has extensive genomic resources, including a well-assembled and annotated genome, and a comprehensive transcriptome data set of developmental stages Zhang et al.
In addition, C. Finally, the taxonomic lineage leading to C. We computed the TAI of C. Briefly, we first organized the C. We then determined the TAI value of each developmental stage based on relative expression of genes from different phylostrata using transcriptomic data see Materials and Methods for details.
This indicates that this stage has relatively high expression of younger genes and is thus less conserved than earlier and later stages by this measure. This pattern directly opposes suggestions that the trochophore stage is the phylotypic stage of spiralians based on morphology Cohen and Massey ; Slack ; Shigeno et al. It is also contrary to a study arguing support of the hourglass model based on the expression patterns of novel homeobox genes in this group Paps et al.
Phylotranscriptomic study of Crassostrea gigas. Trochophore stages are indicated with dashed lines. The shaded area represents the standard deviation estimated by permutation analysis.
The hourglass pattern observed in other studies would be relatively higher at early and late stages, and lower at some midembryonic stage, for instance the trochophore. The number of genes assigned to each phylostrata is shown within brackets. The standard deviation and significance test results of B are shown in supplementary figure S2 , Supplementary Material online. We followed most other previous TAI studies based on transcriptomic sequencing data by normalizing the data TPM normalization but not transforming it.
Two recent studies show that square root or log transformation of the expression data can change the observed pattern Piasecka et al. We performed square root transformation to our TAI result and it still shows a significant peak at the trochophore stage, but it is less prominent compared with the pattern calculated by the untransformed data compare supplementary fig.
S1 , Supplementary Material online with fig. The previous report of the TAI profile of C. Their study was performed when only a few spiralian genomes and transcriptomes were available, thus the taxonomic resolution was low; for example, only three phylostrata younger than Spiralia were used: Mollusca, Bivalvia, and C.
We wondered whether inclusion of more data would show that some of the genes that were considered species-specific are actually in older phylostrata. Indeed, in our study, with a more comprehensive spiralian database, there is better taxonomic resolution in spiralians and far fewer genes assigned to the species-specific category fig.
In contrast to the previous study, the reverse hourglass pattern we observed persists even after removing all phylostrata younger than Bivalvia. When the reverse hourglass pattern does disappear, the remaining TAI profile is generally flat—there is no other large peak or trough in the pattern. This result is conservative because with more spiralian genomic data added in the future, more genes will be assigned to older phylostrata.
In sum, the reverse hourglass pattern found in C. The reverse hourglass may have been caused by abnormal development of experimental subject animals or experimental errors during production or assembly of the transcriptome data. To address this concern, we performed an additional TAI calculation based on a recently published experimental replicate of C.
As shown in figure 2C , the TAI profile of this replicated experiment shows a similar result to our first set of experimental data, with the highest TAI in the trochophore stage. Thus, this pattern is unlikely to be caused by experimental variation. A difference in the computation of TAI also contributed to the discrepancy between our results and those of Xu et al.
TAI is calculated as the phylostrata-weighted sum of gene expression divided by the sum of unweighted gene expression. While this may be suitable to show the contribution of each phylostrata to the overall TAI pattern, it is not appropriate for studying the developmental TAI profile after certain phylostrata are removed because it will systematically bias the results, in some cases reversing the pattern.
For instance, if the TAI of the peak stage was largely contributed by younger phylostrata, when only removing the weighted sum of gene expression from these younger phylostrata from the numerator, the TAI of the peak stage will be reduced to the lowest. This is because the sum of the gene expression the denominator of the peak stage represents the largest of all stages, as this stage has the highest expression of those younger phylostrata. This explains why they reported a reverse hourglass for the overall pattern, but an hourglass pattern after removing the younger genes.
In our study, we removed gene expression from both the numerator and the denominator when we removed successive phylostrata; this resulted in a reverse-hourglass model even after removal of phylostrata younger than Bivalvia see Materials and Methods for details.
When we employed the calculation method from other previous analyses for our data, we observed the hourglass pattern after removing phylostrata younger than Mollusc fig.
The TAI profiles of Crassostrea gigas after removing phylostrata younger than Mollusca, using two different calculation methods applied to the same expression data set Zhang et al. A TAI profiles based on phylostrata assignment and gene expression analysis in this study, using data from Zhang et al. B TAI profiles based on phylostrata assignment and gene expression analysis from Xu et al.
There are fewer timepoints than in A because they used a subset of the stages in Zhang et al. In both panels: the top line is the TAI profile using all phylostrata; the middle line is the TAI profile after removing phylostrata younger than Mollusca calculated by the method used in this study; and the bottom line is the TAI profile after removing phylostrata younger than Mollusca, calculated by the method used by previous studies, including Xu et al.
Abbreviations follow those in figure 2. Another method to measure relative conservation of developmental stages is to compare the expression of conserved and fast-evolving genes. This approach generates what is called the transcriptome divergence index TDI; Quint et al. Compared with TAI analysis, which assigns genes to different phylostrata, the TDI uses the sequence divergence of genes to represent the evolutionary conservation of a gene.
As with the TAI, the distance is weighted by gene expression. A higher TDI value indicates higher expression of fast-evolving genes, or less evolutionary constraint within a stage. TDI reflects the selection pressure on the development stage. Since TDI also indicates whether the observed pattern is actively maintained Drost et al. The TDI profiles of Crassostrea gigas development.
Abbreviations are the same as figure 2. To determine if the reverse hourglass pattern is restricted to C. The TAI profile of the abalone H. Thus, the profile for this species generally resembles a reverse hourglass but it is less pronounced than for the C.
The profile of the annelid P. These findings indicate that the reverse hourglass pattern is not restricted to oyster C. Both have their highest TAI in one of the trochophore stages. Here, we report a reverse hourglass pattern in spiralian development using phylotranscriptomic approaches in two species of molluscs and one annelid.
In the oyster C. Our results indicate that, contrary to earlier conclusions based on comparative morphology Cohen and Massey ; Slack ; Shigeno et al. This result seems to be true for another mollusc and an annelid. The trochophore stage occurs during organogenesis in spiralians.
It has been argued that phylotypic stages in other taxa occur during organogenesis because different organ systems have to coordinate development resulting in a complex pattern of interactions Raff It is possible that spiralian organogenesis is less integrated than in other clades.
Given this view, it could be argued that the apparent lack of constraint during spiralian organogenesis reflects a lack of signaling between clones that are making different organs.
However, the available evidence suggests that cell signaling is important in spiralian embryos, indicating that lineages are not independently developing in organogenesis. Cases of signaling and regulation during organogenesis are known Cather ; Chan and Lambert , and several organs are formed from combinations of cells from different clones, which likely requires communication between cell populations e.
Thus, we would expect that the developmental integration of various lineages in spiralian organogenesis is similar to other taxa, and unlikely to explain the reverse hourglass pattern. In other groups, the Hox genes are patterning the embryo during the phylotypic stage, and regulatory interactions between these genes have been invoked to explain the hourglass model Duboule S18 in Zhang et al.
Overall, three Hox genes have peak expression earlier than the trochophore stage, two have peak expression in the trochophore, and five have peak expression after. Thus, the trochophore stage does not appear to be a particularly important stage for Hox gene patterning. It may be that the pattern we observe is caused by more specific processes that are occurring in the trochophore stage.
In fact, some key spiralian and molluscan synapomorphies are developing at this time. Ectomesoderm is a spiralian-specific form of mesoderm that is proliferating and forming body muscles at this stage, so genes that are specifically associated with ectomesoderm could contribute to the pattern.
The trochophore is named after the primary ciliary band of the larva, the prototroch. This structure appears near the beginning of trochophore stage, and is elaborated progressively in the groups considered here. If ciliary band genes are relatively young and fast-evolving, they could also be contributing. The shell is a molluscan synapomorphy which starts to develop shortly before the trochophore stage. Intriguingly, biomineralization proteins that control shell deposition are enriched for fast-evolving and novel genes Aguilera et al.
One thing I've learned about scientific work is that reading papers makes it seem easy, at least on paper excuse the pun This blog tells the story behind the project, the tough moments and the challenges. I remember the day we came up with the idea for this project. It was right after a group meeting, where we discussed a recently published paper by Mirko Fracesconi and Ben Lehner Nature , In that paper, the authors showed how developmental gene expression patterns change due to genetic variations.
Being in a lab with a great interest in the evolution of development evo-devo , we wondered how we can use such an approach to tackle an age old question - why is there an embryonic stage that is much more conserved than the rest across species of the same phylum? I am referring, of course, to the phylotypic stage, whose increased conservation is summarized by the hourglass model of development.
We recognized that studying the pattern of gene expression differences throughout development would allow us to search for evidence for this mythical stage.
I was very excited that in this post-genomic era we have the capability to answer old questions with new molecular tools. The approach that we came up with was to compare gene expression of all genes among strains, across embryonic time, and do it with C. We figured that different strains would serve as a good model for divergence within a species, and the question would be — considering the variations between the genomes of different strains, how will different embryonic stages be affected?
Will the phylotypic stage remain unchanged? We initially considered using the same strains as in Francesconi and Lehner's paper, but these were not a good model for divergence as these strains were made by crossing two C. We wanted strains with mutations that could not even be attributed to positive selection. Instead, we realized that mutation-accumulation strains were perfect for our goal.
They are all derived from a common ancestor, possessing myriad mutations, and they are easy enough to work with. That was the easy part. When the fire of the wet lab work settled, I was standing in front of the ruins I created — big data!
What am I supposed to do with it now? I had never coded before, and getting into it gave me a sensation of superiority: "Look at you, big data! You're not so scary! We tried all kinds of methods described by others in the field, but none seemed to be appropriate for our data.
0コメント