Advanced Biological Data Analysis

Vakinformatie

Keuze-truncus vak voor master Biochemie (verplicht voor master Biologie), gegeven door Piet van den Berg, Tom Wenseleers & Hans Jacquemyn. Open boek examen waar je 2 datasets moet analyseren die gelijkaardig zijn aan die van de oefenzittingen. 10/20 voor het onderdeel van PVDB & TW. 10/20 voor het onderdeel van HJ.

Dit werd vroeger gegeven door professors Tom Wenseleers en Hans Jacquemyn. Het examen is schriftelijk open boek, en bestaat uit het ter plekke statistisch analyseren van verschillende datasets op de computer. Hierbij staat het deel gegeven door Wenseleers en Jacquemyn elk op 10 van de 20 punten.

ECTS-fiche: https://onderwijsaanbod.kuleuven.be/syllabi/e/G0F87AE.htm

Examenvragen

14 januari 2025

Piet van den Berg & Tom Wenseleers

The brown skua (Catharacta antarctica) is a large predatory seabird that occurs in the Antarctic and subantarctic regions. The species exhibits sexual dimorphism that goes in the opposite direction than what is typically observed: females are about 5% larger than males. In this study, the researchers aimed to investigate whether there are differences in cognitive performance between the sexes of this species. To measure this, they developed a ‘string-pulling task’ in which the birds were presented with two strings, only one of which was attached to a reward. The birds were allowed to pull only one of the strings to try to obtain the reward. The entire string and the reward were visible to the birds, so it was in principle possible for them to work out which string would yield the reward. There were two conditions: a ‘parallel’ condition, in which the two strings ran in parallel next to each other, and a (more difficult) ‘crossed’ condition, where the strings crossed and so it was harder to see which string would yield the reward. They tested 11 individual birds (6 females and 5 males, individual birds are labelled in the column ‘Bird_ID’) for 16 consecutive trials in both experimental conditions (so a total of 32 trials per bird). In each trial, they recorded whether the bird pulled the string that had the reward (‘Success’ = 1) or the string without the reward (‘Success’ = 0). They randomized which string had the reward between trials.

Run a data analysis in R to answer the question below. Then write your answer on the next page of this document. Make sure your answer is clear and concise, and that your conclusion is supported by the data analysis that you performed. Paste the output (tables and figures) that is needed to support your argument to this document, but don’t paste code in this file (hand in your R script separately and make sure that it is clear). Make sure that your name is in the filename of both documents!

QUESTION: Are there any differences in cognitive ability between male and female birds across the different experimental conditions?

Run models to answer the question above. At least include Sex and experimental Condition as predictor variables in each model (since these are explicitly relevant to the research question). Also consider models that include the interaction between these two variables and models that include Trial, but don’t include interactions between Trial and the other variables. Include the appropriate random effects structure in each model. Check for overdispersion. Base your conclusions on a proper model selection procedure. For the purpose of this exercise, there is no need to do any other model diagnosis checks. Visualize the best model with an as good as possible visual representation. Write a clear conclusion (important!).

(Excel contained Bird_ID, Sex, Condition, Succes and Trial)

Hans Jacquemyn

Data set: Epipactis.xlsx

Ecological processes are central to the formation of new species. When barriers to gene flow emerge between populations (e.g. shifts in flowering time, landscape changes), there is an increased likelihood that divergent selection and local adaptation will lead to the rapid formation of ecotypes, and potentially new species. However, it is often not clear to what extent populations have diverged genetically and morphologically. This may be particularly true for the orchid genus Epipactis. This genus contains a problematical complex of taxa among which species limits are difficult to define. As a result, different authors have treated the taxonomy of Epipactis in different ways, some recognizing the different taxa as distinct species, others considering them as minor intraspecific variants. In dune habitats in Belgium, plants that closely resemble the widespread Epipactis helleborine can be regularly encountered. The precise taxonomic position of these plants, however, is not clear and currently subject to debate; some regard it as a variety of E. helleborine (E. helleborine var. neerlandica), some as a subspecies (E. helleborine subsp. neerlandica), whereas others consider it as a distinct species (E. neerlandica). In this study, 28 different morphological characters were measured for plants growing in two dune (WH2 and TY1) and two forest (MW1 and MW2) populations to investigate to what extent dune and forest ecotypes differ morphologically. These characters were related to the size of the plant (traits T1–T14), leaf characteristics (T15–T16), and floral morphology (T17–T28).

1) Investigate whether the overall morphology differs significantly between plants of the forest and dune ecotype.

2) Illustrate graphically which variables best explain variation in morphology between both ecotypes and identify the variables that most determine the morphological difference between the two ecotypes.

3) Can you notice a significant difference in overall morphology between the different populations that were sampled and did these differences depend on the ecotype?

(Excel contained population, ecotype, plant and T1-28 (continuous data))

2024

Piet van den Berg & Tom Wenseleers

Clownfish (Amphiprion spp.) are marine fish found in reefs in the eastern Indian Ocean and the western Pacific that have symbiotic relationships with sea anemones. They live in highly socially structured harems and are aggressively territorial.

In this study, the researchers were interested in understanding the aggressive behaviours of two species of clownfish: - the common clownfish (A. ocellaris) and - the orange clownfish (A. percula; variable ‘AmphiprionSpecies’ in the dataset)

They observed aggressive behaviours towards unfamiliar conspecifics of the fish at 24 different territories (indicated with ‘TerritoryID’ in the dataset), which each housed one or multiple clownfish harems. The researchers presented various males and females (variable ‘Sex’) from each territory with an unfamiliar conspecific, and then counted the number of aggressive behaviours that the fish exhibited in a timeframe of 5 minutes (‘AgressionEvents’). Run a data analysis in R to answer the question below. Then write your answer on the next page of this document. Make sure your answer is clear and concise, and that your conclusion is supported by the data analysis that you performed. Paste the output (tables and figures) that is needed to support your argument to this document, but don’t paste code in this file (hand in your R script separately and make sure that it is clear). Make sure that your name is in the filename of both documents!   QUESTION: How does aggression depend on species and sex? - Run models that consider all combinations of variables mentioned in the question above (including models that consider only single predictors). - Also run a model that considers the interaction. - Choose the best model. - Check for overdispersion and continue with the appropriate model. - For the purpose of this exercise, there is no need to do any other model diagnosis checks. - Visualize the best model with an as good as possible visual representation. Write a clear conclusion (important!).

Hans Jacquemyn

Ecological processes are central to the formation of new species. When barriers to gene flow emerge between populations (e.g. shifts in flowering time, landscape changes), there is an increased likelihood that divergent selection and local adaptation will lead to the rapid formation of ecotypes, and potentially new species. However, it is often not clear to what extent populations have diverged genetically and morphologically. This may be particularly true for the orchid genus Epipactis. This genus contains a problematical complex of taxa among which species limits are difficult to define. As a result, different authors have treated the taxonomy of Epipactis in different ways, some recognizing the different taxa as distinct species, others considering them as minor intraspecific variants. In dune habitats in Belgium, plants that closely resemble the widespread Epipactis helleborine can be regularly encountered. The precise taxonomic position of these plants, however, is not clear and currently subject to debate; some regard it as a variety of E. helleborine (E. helleborine var. neerlandica), some as a subspecies (E. helleborine subsp. neerlandica), whereas others consider it as a distinct species (E. neerlandica). In this study, 28 different morphological characters were measured for plants growing in dune and forest habitats to investigate to what extent dune and forest ecotypes differ morphologically. These characters were related to the size of the plant (traits T1–T14), leaf characteristics (T15–T16), and floral morphology (T17–T28).

1) Investigate whether the overall morphology differs significantly between plants of the forest and dune ecotype and identify the variables that most determine the morphological difference between the two ecotypes.

2) Illustrate graphically which variables best explain variation in morphology between both ecotypes.

3) Can you notice a significant difference in overall morphology between the different populations that were sampled and did these differences depend on the ecotype? Illustrate your answer with the appropriate graph.

2024

Piet van den Berg & Tom Wenseleers

Swifts are insectivorous birds belonging to family of Apodidae. Swifts resemble swallows in both morphology and niche, but the two families are only distantly related – their similarity is due to convergent evolution. They occur on all continents and those endemic to temperate regions are typically migratory. Swifts are particularly active during twilight (when their insect prey also tend to be more active) and can ascend to high altitudes when they are foraging.

In this study, the researchers wanted to know if swifts are influenced by the brightness of the moon during their foraging bouts. Specifically, they were interested to know if moon illuminance affected the altitude reached during their foraging bouts by swifts of three species: the common swift (Apus apus), the pallid swift (Apus pallidus) and the alpine swift (Tachymarptis melba). The researchers equipped the animals with sensors that allow to make rough altitude estimations. For a large number of foraging bouts, the researchers logged whether the bird reached a high altitude or not (‘high’ was defined as more than the mean plus a standard deviation of flight altitudes during the day; this is recorded under the variable “altbin” in the dataset: 1 for a high altitude and 0 for a low altitude). They also measured moon illuminance (“moon_illuminance”) during each foraging bout and included the species of the measured individual in the dataset (“species”). They had many measurements for each individual (the individual is indicated by “tag” in the dataset). Run a data analysis in R to answer the question below. Then write your answer on the next page of this document. Make sure your answer is clear and concise, and that your conclusion is supported by the data analysis that you performed. Paste the output (tables and figures) that is needed to support your argument to this document, but don’t paste code in this file (hand in your R script separately and make sure that it is clear). Make sure that your name is in the filename of both documents! Exercise QUESTION: How does moon illuminance affect flight altitude in the three species of swift? - Run models that consider all combinations of variables mentioned above (including models that consider only single predictors).

- Also run a model that considers the interaction. Choose the best model. - Check for overdispersion and continue with the appropriate model. - For the purpose of this exercise, there is no need to do any other model diagnosis checks. Visualize the best model with an as good as possible visual representation. - Write a clear conclusion (important!).

Hans Jacquemyn

Primula veris is a distylous plant species that can be found in forest and grassland habitats (Fig. 1). Populations of P. veris typically contain long-styled (L-morph) and short-styled (S-morph) plants. Due to the pronounced differences in environmental conditions (e.g. light conditions, soil moisture content) between forests and grasslands, we can expect that plants growing in these habitats display different growth characteristics.

Fig. 1 Example of Primula veris growing in grassland habitat.

To test this hypothesis, a large number of plant characteristics were measured: • Leaf dry mass (Leaf_dry_mass) (g) • Number of leafs (Leafs) • Number of flower stalks (Flower_stalks) • Average length of flowering stalks (Stalk_length) (cm) • Total number of flowers (Total_flowers) • Average number of flowers per flowering stalk (Average_flowers) • Leaf surface area (Leaf_surface_area) (cm²) • Leaf width (Leaf_width) (cm) • Leaf length (Leaf_length) (cm) • Number of stomata (Stomata) • Specific leaf area (SLA) : ratio of leaf area to dry mass

Besides, information on the habitat (forest or grassland), morph type (short-styled (S) or long-styled (L)), and the population from which plants were sampled (Population) is provided as well.

1) Investigate whether the overall morphology differs significantly between plants from grasslands and plants in forests and whether this difference is significantly affected by morph type.

2) Which variables are most important in determining the difference in overall morphology between forest and grassland plants?

Illustrate your results with the appropriate figures. Please also explain in detail why you have chosen for a particular analysis technique.

30 januari 2023

Piet van den Berg & Tom Wenseleers

In this study, the researchers were interested in how dominance relationships affect food sharing in the wild. In 4,549 hours of observations, they recorded all instances where there was the possibility for sharing food (potential sharing events). This was the case when one individual had food (the “possessor”) and another individual begged to obtain some of it (the “potential partner”). They recorded whether the begging behaviour resulted in the possessor sharing food with the potential partner (“Sharing” in the dataset; 1 if food was shared, 0 if not). They gave each individual an identity tag and recorded these for both individuals in each potential sharing event (“Possessor” and “PotentialPartner” in the dataset) and they also recorded their dominance ranks (“DominancePossessor” and “DominancePotentialPartner”). Most individuals were observed in multiple potential sharing events.

QUESTION: How does dominance rank affect food sharing in chimpanzees?

Run models with the dominance ranks of the possessor and the potential partner as predictors. Consider all combinations (both ranks separately, both together, and a model that also includes the interaction). Choose the best model. Include the appropriate random effects structure in all models that you run. Check for overdispersion and continue with the appropriate model. For the purpose of this exercise, there is no need to do any other model diagnosis checks. Visualize the best model. Write a clear conclusion (important!).

Hans Jacquemyn

1. The marsh orchid (Dactylorhiza sphagnicola) and the heat-spotted orchid (Dactylorhiza maculata) (Fig. 1) are two orchid species that are able to hybridize when they co-occur. In the Belgian Ardennes, both allopatric and sympatric populations of both orchids can be found and there are some indications that hybridization occurs in sympatric populations. In order to get better insights in the hybridization process, the morphology of a large number of plants was investigated in a large sympatric population where both species co-occurred. In total 28 morphological traits were measured (Table 1). The same characteristics were also measured for a large number of plants in allopatric populations (one for each species). The data have been summarized in two datasets. The dataset sympatric.xlsx contains all data for the sympatric population, the dataset allopatric.xlsx contains the data for the allopatric populations.
   • A. Investigate whether the two species can be unequivocally distinguished based on the measured characteristics?
   • B. Which traits are most important to distinguish the two species?
   • C. Are there indications that hybridization has occurred in the sympatric population? How can you deduce this from your analysis?

17 januari 2020 NM

Tom Wenseleers - oral (you receive the questions in an R file)

1. The provided data file contains detailed data on the survival of the passengers of the titanic (survived=0/1 for passengers that died or not) as a function of, amongst others, their age ("age"), their sex ("sex") and the price they paid for their ticket ("priceperticket").
   • A. Fit a model of passenger survival in function of age, sex & priceperticket with the appropriate error structure. Start with a model that takes into account all possible higher-order interaction effects and then use stepwise backward model reduction and calculate the best model based on the bayesian information criterion bic. Test for the possible presence of outliers & influential observations and remove those if necessary.
   • B. Make effect plots of all the predictors in your best model (on the link scale but using y axis response plot labels, using type="rescale") and interpret the results.
   • C. Calculate the log(odds ratio) to survive for male passengers of average age that paid for the most expensive ticket vs those that paid for the cheapest one. Do the same for female passengers. Did the log odds of them surviving go up in proportion to the price they paid for their ticket, or was there an unfair advantage for those that paid for the most expensive one?

Paste all R code + all tabular & graphical output & interpretation in a word document and do the same for the part of hans jacquemyn and hand it over to the assistant saved as lastname_firstname.docx & lastname_firstname.r (zip these two files as lastname_firstname_exam.zip)

Hans Jacquemyn (in a Word document)

1. Orchids rely on mycorrhizal fungi to complete their life cycle. Recent research has shown that these fungi can be quite diverse and belong to different genera. Fungal communities are also likely to vary between orchid populations. Little,however, is known about the factors determining variation in fungal communities and whether this variation affects the population dynamics of orchids. It can be expected that differences in soil conditions have a significant impact on mycorrhizal communities and therefore impact on the population dynamics of the orchids. To test this hypothesis fungal communities were determined in a large number of populations of the terrestrial orchid Neottia ovata (Fig. 1). For each population a series of soil characteristics was measured and the change in population size was assessed by comparing the population size in 2003 with that in 2013. Data were collected in three separate plots in each population. The data have been summarized in two datasets, one that gives for each population the abundance (number of sequences) of all detected fungi (Neottia_fungi.xlsx), and one that gives an overview of the soil characteristics (Neottia_soil.xlsx).
   • The soil characteristics that were measured are soil moisture content (moist), organic matter (OM), phosphate concentration (P), nitrate concentration (NO3) and ammonium concentration (NH4). The last file also tells you whether a population has increased (1) or decreased in size (0) between 2003 and 2013.
   • A. Assess whether fungal communities vary among populations?
   • B. Investigate whether variation in fungal communities can be related to soil conditions. Which soil variables have the largest effect on fungal communities?
   • C. Investigate whether the observed changes in population size can be related to variation in mycorrhizal communities.

17 januari 2020 VM

Tom Wenseleers - oral (you receive the questions in an R file)

1. Researchers were interested to test if queen pheromones of the honeybee, which in that species are emitted by the queen to stop the workers from reproducing, would also inhibit the reproduction of either workers or queens in the bumblebee. They hypothesized that such cross activity might be observed in the event that these compounds exploit conserved physiological pathways linked with the regulation of reproduction. To test this hypothesis, the researchers exposed bumblebee queens and bumblebee worker groups (groups of 20 workers each) to either a blank solvent-only control or a solution of honeybee queen pheromones for a period of 2 weeks. Subsequently, they dissected the queens and workers and measured the size of the largest oocyte in their ovaries to be able to test for any effects on ovary development. For each of the worker groups, these measurements were averaged over all individuals. In terms of experimental design, genetic background was controlled for by doing the experiment in a paired fashion, whereby the worker groups exposed to each replicate control and queen pheromone treatment were derived from the same source colony and similarly, the queens exposed to each replicate control and queen pheromone treatment were taken to be sisters of each other (colony or sibgroup is encoded as variable "id" in the dataset). In your analysis, take into account this non-independence through the inclusion of a random effect term, and use a model with the appropriate error distribution.
   • Aim: Test whether honeybee queen pheromones inhibit ovary development and whether this effect is caste dependent. (cf. dataset "data.csv").
   • SPECIFIC QUESTIONS:
     • A. Display your data using "spaghetti plots", i.e. plot oocyte size in function of treatment, using two different panels for caste, and connect points that are measured from individuals from the same colony (for workers) or sib-group (for queens). Make this plot both using lattice's xyplot and ggplot2.
     • B. Fit a model of oocyte size (SIZE_OOCYTE) in function of TREATMENT and CASTE, either considering a possible interaction effect between both or not, taking into account possible random effects and use a model with the appropriate error distribution. Decide which model is best based on the AIC criterion. What is the name of the type of model you fitted? (in this case a linear mixed effects model was ok, since the distribution or errors was normal: lme() or lmer())
     • C. Make effect plots of the effects in your model and explain what these imply.
     • D. Carry out the relevant tests for the significance of the different effects. What do these tell you? Also carry out Tukey posthoc tests to test the effect of treatment for each of the two castes. What would the conclusion have been if you would have ignored the dependency in your data (variable ID, i.e. colony or sibship)? WOuld the effect of TREATMENT have been more or less significant then and why? (just do a normal lm here)
     • E. Test whether the residuals of your model conform to your assumed error distribution.

Hans Jacquemyn (in a Word document)

1. The marsh orchid (Dactylorhiza sphagnicola) and the heat-spotted orchid (Dactylorhiza maculata) (Fig. 1) are two orchid species that are able to hybridize when they co-occur. In the Belgian Ardennes, both allopatric and sympatric populations of both orchids can be found and there are some indications that hybridization occurs in sympatric populations. In order to get better insights in the hybridization process, the morphology of a large number of plants was investigated in a large sympatric population where both species co-occurred. In total 28 morphological traits were measured (Table 1). The same characteristics were also measured for a large number of plants in allopatric populations (one for each species). The data have been summarized in two datasets. The dataset sympatric.xlsx contains all data for the sympatric population, the dataset allopatric.xlsx contains the data for the allopatric populations.
   • Given Vegetative traits: Plant height from soil level (cm), Number of cauline leaves, Lowermost leaf length (cm), Lowermost leaf maximum width (cm), Length of second leaf (cm), Maximum width of the second leaf (cm), Position of the second leaf greatest width (cm), Uppermost leaf length (cm), Uppermost internodium length (cm), Stem diameter under infl orescence (mm), Stem diameter above lowermost leaf (mm), Number of flowers, Inflorescence length (cm), Length of inflorescence axis between the bract insertion points of first and fifth flowers (cm)
   • Given Flower traits: Bract length (cm), Bract width (cm), Ovary length (cm), Lateral sepals length (cm), Lateral sepals width (cm), Petals length (cm), Petals width (cm), Labellum length (cm), Labellum lateral lobes length from base (cm), Labellum median lobe length (cm), Labellum width (cm), Labellum median lobe width at base (cm), Spur length (cm), Spur diameter at base (cm)
   • A. Investigate whether the two species can be unequivocally distinguished based on the measured characteristics?
   • B. Which traits are most important to distinguish the two species?
   • C. Are there indications that hybridization has occurred in the sympatric population? How can you deduce this from your analysis? Illustrate your answer with the appropriate figures and explain which analysis you have used and why.

6 januari 2019

Tom Wenseleers - oral (you receive the questions in an R file)

1. Researchers were interested to test if queen pheromones of the honeybee, which in that species are emitted by the queen to stop the workers from reproducing, would also inhibit the reproduction of either workers or queens in the bumblebee. They hypothesized that such cross activity might be observed in the event that these compounds exploit conserved physiological pathways linked with the regulation of reproduction. To test this hypothesis, the researchers exposed bumblebee queens and bumblebee worker groups (groups of 20 workers each) to either a blank solvent-only control or a solution of honeybee queen pheromones for a period of 2 weeks. Subsequently, they dissected the queens and workers and measured the size of the largest oocyte in their ovaries to be able to test for any effects on ovary development. For each of the worker groups, these measurements were averaged over all individuals. In terms of experimental design, genetic background was controlled for by doing the experiment in a paired fashion, whereby the worker groups exposed to each replicate control and queen pheromone treatment were derived from the same source colony and similarly, the queens exposed to each replicate control and queen pheromone treatment were taken to be sisters of each other (colony or sibgroup is encoded as variable "id" in the dataset). In your analysis, take into account this non-independence through the inclusion of a random effect term, and use a model with the appropriate error distribution.
   • Aim: Test whether honeybee queen pheromones inhibit ovary development and whether this effect is caste dependent. (cf. dataset "data.csv").
   • Specific questions:
     • A. Display your data using "spaghetti plots", i.e. plot oocyte size in function of treatment, using two different panels for caste, and connect points that are measured from individuals from the same colony (for workers) or sib-group (for queens). Make this plot both using lattice's xyplot and ggplot2.
     • B. Fit a model of oocyte size (SIZE_OOCYTE) in function of TREATMENT and CASTE, either considering a possible interaction effect between both or not, taking into account possible random effects and use a model with the appropriate error distribution. Decide which model is best based on the AIC criterion. What is the name of the type of model you fitted? (in this case a linear mixed effects model was ok, since the distribution or errors was normal: lme() or lmer())
     • C. Make effect plots of the effects in your model and explain what these imply.
     • D. Carry out the relevant tests for the significance of the different effects. What do these tell you? Also carry out Tukey posthoc tests to test the effect of treatment for each of the two castes. What would the conclusion have been if you would have ignored the dependency in your data (variable ID, i.e. colony or sibship)? WOuld the effect of TREATMENT have been more or less significant then and why? (just do a normal lm here)
     • E. Test whether the residuals of your model conform to your assumed error distribution. (he wanted to see a histogram of the residuals)

Hans Jacquemyn (in a Word document)

1. Evolutionary theory predicts that coexistence of two closely related plant species can have a major impact on floral morphology and plant mating systems. To test this prediction, floral morphology of a large number of individuals from both allopatric and sympatric populations of Common centaury (Centaurium erythraea) and Seaside centaury (C. littorale) (Fig. 1) were investigated along the Belgian coast. For each individual seven floral traits were measured (Table 1). (Data set: Centaurium.xlsx) (Table 1: Floral traits measured in allopatric and sympatric populations of Centaurium erythraea en C. littorale along the Belgian coast. (Given traits and variables were: Total length of the flower (mm), Total Petal length (mm), Petal width (mm), Ovary length (mm), Style length (mm), Length of the anthers (mm), Level of herkogamy (mm))
   • A. Investigate whether floral morphology differs significantly between flowers of Centaurium erythraea and C. littorale.
   • B. Illustrate graphically which variables best explain variation in floral morphology between both species and indicate which variable most determines the difference between the two species.
   • C. Can you notice a difference in floral morphology between allopatric and sympatric populations? Is this difference determined by the identity of the species? If yes, explain how and illustrate with the appropriate graph.
   • D. Based on your results, can you conclude that allopatric populations can as easily discerned from sympatric populations for C. erythraea as for C. littorale? And which floral traits are best suited to discern allopatric from sympatric populations?