During development complex phenotypes are formed without immediate natural selective feedback. I will discuss the crucial role of hierarchy - both in gene expression, signaling and in cell architecture - in promoting local selective processes that stabilize developmental patterning. I will review some of the data on gene knockouts and relate these empirical results to the idea of facilitated variation.
Some classes of RNA molecules, among them certain non-coding RNAs and viral RNA motifs are subject to strong selection on their structure. The biophysical properties of RNA imply a strong correlation between robustness (against mutations) and thermodynamic stability. Hence it is non-trivial to disentangle direct selection for stability from selection for robustness. I will briefly present the underlying models and review the pertinent literature, closing with a few ideas to tackle the problem.
We study the relation between correlation and causation within the setting of Bayesian networks. By using an information flow measure for causation we provide a quantitative extension of Reichenbach's principle of common cause. In particular, we clarify in what sense one can infer causal relations from the correlation of variables. Furthermore, we will discuss the role of general interventions including knockout perturbations for the identification of a system's mechanisms.
In biological systems, highly robust information processing is crucial for fitness and survival. System output must be reproducible despite the intrinsic noise of the elements (genetic switches, neurons, etc.). Such noise poses severe stability problems to parallel information processing as it tends to desynchronize system dynamics (e.g. via fluctuating response or transmission time of the elements). We study the reliability of the output from networks of autonomous noisy elements. We find that the presence or absence of reliable dynamical attractors with self-sustained synchrony strongly depends on the underlying circuitry. Our model suggests that the observed motif structure of biological signaling networks is shaped by the biological requirement for reproducibility of attractors.
A simple versatile statistical learning algorithm for inferring large-scale causal networks is proposed. The basis of this procedure is a decomposition of the regression coefficient into the product of the square root of the ratio of standardized partial variances, the partial correlation, and a scale factor. Multiple testing of the log-ratio of standardized partial variances produces an estimate of the partial ordering of the nodes, which is subsequently projected on the underlying partial correlation graph. As a result, a (partially) causal network is recovered, using only the positive definite pairwise correlation matrix as input. The statistical and computational properties of the new approach are investigated, and illustrated by analyzing a large-scale Arabidopis thaliana gene expression time series experiment.
