Fish Parasite Project

Created on 06-02-2023

Table of Contents

1. About The Project
   • Key File Description
   • Statistical Hypotheses
2. Getting Started
   • Prerequisites
3. Contact

About The Project

This project looks at the effect of internal parasites on the fish host personality, the pumpkinseed sunfish. Data were collected between May-September 2022 at la Station de Biologie des Laurentides, Université de Montréal, Canada.

The objective is to determine whether an experimental co-infection caused by trematodes (Uvulifer ambloplitis and Apophallus sp) and cestodes (Proteocephalus ambloplitis) will change the host’s personality (Lepomis gibbosus). We looked at exploration, boldness and activity before and after an experimental infection in a semi-natural environment (caging experiment in lake).

This project will generate complex datasets (i.e. repeated measures of behavioral traits through time, responses following a perturbation). This type of data structure could apply to researchers working on a wide range of physiological, morphological or behavioural traits.

Key File Description

1. In the main folder of the project fish_parasite, we find:
   a) fish_parasite.Rproj: the R project
   b) results.qmd : all the written results and the figures/tables

2. Data_raw folder contains:
   a) all_data.csv : all the raw data collected for the Fish Parasite Project
   b) meta.data.csv : all the variables explained with unit of measurement
   c) 12_fish.csv : some data for 12 fish that were excluded from the analysis

3. R folder contains:
   b) fish_parasite.R : the R script used in the project
   c) func : functions used in the scripts
   d) processing_data : all the processing/transformations for the raw data
   e) model : we have model 1 to model 10, which are all the models we did during the analysis

4. output folder contains:
   a) all_data_p : all the processing data
   b) dat_models : data used for models using all fish
   c) dat_models_C : data used for models using only control (or uninfected fish)
   d) dat_models_E : data used for models using only experimentally infected fish
   e) dat_models_parasite : data used for models looking at parasites
   f) dat_slopes : processed data for model 3
   g) new_data : processed data for model 10
   h) new_data1 : merge of new_data and dat_slopes for model 10

• We also have a folder with all the models and figures output

Biological Hypotheses

Analysis Step 1:

Question 1: Do we find evidence of personality (i.e., consistent differences in behaviour among individuals)? Does boldness, exploration and activity form a beavioural syndrome?

Question 2: Does parasitic infection impact the repeatability of each trait and the strength of behavioural syndromes?

• Hypothesis 1: We expect that the repeatability of the traits to be reduced because parasites should reduce between individual variance to maximize transmission. Behavioural syndromes should be un-affected.

• Hypothesis 2: We expect that the repeatability of the traits to increase because different parasite and host genotypes / immune systems will result in some parasites being more or less affected which should increase between individual variance. Behavioural syndromes might be expected to become stronger (more correlated).

Analysis Step 2:

Question 2: How does behaviour change with parasite infection?

Hypothesis 1: Parasites increase boldness, exploration, and activity because fish are intermediate hosts and increased boldness, activity etc results in a higher probability of predation which facilitates parasite host transmission.

• Prediction 1(H1): We expect mean boldness, exploration and activity to increase in indiviudals experimentally infected with parasites.

Hypothesis 2: Parasites decrease boldness, exploration, and activity because it's energetically costly to mount an immune response against parasite infection.

• Prediction 1(H1): We expect mean boldness, exploration and activity to decrease in indiviudals experimentally infected with parasites.

Statistical hypotheses

Analysis Step 1:

Behavioural syndrome
H0: traits are not correlated
H1: traits are positively correlated (which is often seen with these three traits)

Repeatability
H0: traits remain constant before and after infection (facilitates predictability of behaviors for future infection)
H1: traits do not remain constant before and after infection (infection changes trait constancy, making fish more vulnerable)

Analysis Step 2:

Exploration
H0: experimental infection has no effect on individuals' exploration
H1: experimental infection increases exploration (parasitic manipulation, more risk of predation)
H2: experimental infection decreases exploration (pathological response, amorphous)

Boldness
H0: experimental infection has no effect on boldness
H1: experimental infection increases fish boldness (parasitic manipulation to transmit the parasite, thus more risk of predation by increasing boldness)
H2: experimental infection decreases fish boldness (disease behaviour that makes the fish more cautious for survival)

Activity
H0: experimental infection has no effect on fish activity
H1: experimental infection increases activity (to counteract weakening by infection and seek more resources)
H2: experimental infection decreases activity (since fish are weakened by parasites)

Models:

Three-step Strategy:

Step 1

• Using all data (60 fish and 4 measurements / fish) we will fit the following models:

1. Model 1: [B, E, A] = u + trtment_{E} + tank + (-1 + trtment_{E}| ID) + (1|Cage)
2. Model 2: [B, E, A] = u + trtment_{E} + (-1 + trtment_{E}| ID) + (1|Cage)

• Above models allow us to 1) estimate repeatability for ALL traits; 2) estimate the behavioual trait correlations; 3) estimate these within EACH treatment group (C vs E).

Step 2

• Subset the experimental and control fish into two datasets (60 fish and 3 measurements for each C and E group) then fit the following models:

1. Model 1 (Experimental Group): [B, E, A] = u + z_body_condition + z_parasite_load + tank + (1 | ID) + + (1|Cage)
2. Model 2 (Experimental Group): [B, E, A] = u + z_body_condition + z_parasite_load + z_parasite_load^2 + tank + (1 | ID) + + (1|Cage)
3. Model 3 (Control Group): [B, E, A] = u + z_body_condition + tank + (1 | ID) + (1|Cage)

Step 3

• We have two different parasites in E group. So, we want to fit models that possibly look at interactive effects of the two, but this will depend on how many parameters that we need to estimate above.
• Yes, we want to do this as the parasites have different final hosts and should be in conflict.

Getting Started

How to set the project...

Prerequisites

Packages needed to run the models

pacman::p_load(lme4, rstan, StanHeaders, jsonlite, rstantools, brms, Rcpp, dplyr, here, flextable, pander)

Packages needed for the results (figures,tables...)

pacman::p_load(tidyverse, brms, glmmTMB, gt, latex2exp, posterior, gt, glue, dplyr,magrittr, ggplot2, cowplot, jpeg, magick)

Contact

Maryane Gradito - @mary_gradito - email: maryane.gradito@umontreal.ca

Project Link: https://github.com/MaryaneGradito/fish_parasite

(back to top)

<