1. Modeling the unobservable developmental stability using a Bayesian latent variable model Stefan van Dongen Dept. Biology, University of Antwerp

2. OUTLINE Introduce biological problem –developmental stability Develop statistical model Show some simulation results Interesting results in humans –keep paying attention this is important

3. We develop from 1 cell to what we are now During development mistakes occur due to random noise (DN) –Fluctuations in concentrations –Somatic mutations –Death of cells There are mechanisms that correct for these mistakes: DEVELOPMENTAL STABILITY What is developmental stability?

4. Our genotype and environment in which we live determine our size and shape –unknown in most cases BUT there is some degree of stochasticity PROBLEM: how to estimate this stochastic contribution which has two components (noise and stability) SOLUTION: Look at symmetric traits

5. How to estimate developmental stability? Left and right side develop often under exactly the same environmental conditions (but e.g. handedness) and obviously share the same genotype In the absence of noise they should develop to the same size or shape Noise will cause asymmetry Stability will counteract this effect

6. How to estimate developmental stability? Measure asymmetry Asymmetry estimates the joint action of noise and stability

7. The picture is not so clear Directional asymmetry or antisymmetry: adaptive and genetically determined Gynandromorphs NO STRESS EFFECT

8. The picture is not so clear Subtle forms of asymmetry Early action of stress Continuous scale Phenodeviants & deformations Under severe stress only Binary

9. Why study developmental stability? Charles Darwin EVOLUTIONCONSERVATION BIOLOGY FITNESS (reproductive success, survival, ….) Difficult to measure in field May be related to stability High stability => sufficient energy to achieve high fitness

10. Why study developmental stability? Charles Darwin EVOLUTIONCONSERVATION BIOLOGY FITNESS (reproductive success, survival, ….) Difficult to measure in field May be related to stability High stability => sufficient energy to achieve high fitness INDIVIDUAL LEVEL

11. 1 2 3

12. Asymmetries are usually small: take bias due to measurement error into account measure[i,j]~N(0,  2 ME ) DIRECTIONAL ASYMMETRY FLUCTUATING ASYMMETRY muind[i,j]=interc[i]+side[i,j]  unb_FA[i] where: interc[i]~N(0,  2 ) and: unb_FA[i]~N(0,  2 FA [i]) expected[i,j]=intercept+slope  side[i,j] ERROR

14. unb_FA[i]~N(0,DI[i]) Assume DN to be constant DS[i]~beta(  1,  2 ) DI[i]=DN  (1-DS[i]) The association between asymmetry and stability

15. N=500, 2500 and 5000 DN constant, DS beta-distr. ‘good’ estimate of DS overestimate of DN ‘worse’ estimate of DS ‘good’ estimate of DN ‘failure’ to estimate bimodal pattern of DS The association between asymmetry and stability

17. fitness[i]=interfit+slopefit  DS[i] The association between asymmetry and fitness

18. Further developments Multiple-trait analyses robustness against deviations of model assumptions (normality vs. log-normality) Include variation in DN (stochastic or constant)

19. Interesting results in humans Symmetric persons have higher IQ Criminals have higher asymmetry Symmetric males and females are more attractive Symmetric males have more sexual partners in their life Symmetric males are ‘better lovers’ Female breast asymmetry decreases around period of ovulation Females are more selective for symmetric males around this ovulation period …..

20. One case study