Nature Genetics Review–Special issue about modeling in biology

Modelling

An abundance of data and unprecedented computational power are allowing sophisticated biological models to be devised and tested. This series of articles examines how the coupling of genetics with disciplines such as engineering, statistics, physics and computational biology has enriched our understanding in areas that range from developmental patterning to genetic association analyses.

---

2009

September 2009 Volume 10 No 9

Microfluidic devices for measuring gene network dynamics in single cells

Matthew R. Bennett & Jeff Hasty

<!–p246 | –>doi:10.1038/nrg2625

• Abstract
• Full text
• PDF (661 KB)

August 2009 Volume 10 No 8

Quantitative approaches in developmental biology

Andrew C. Oates, Nicole Gorfinkiel, Marcos González–Gaitán & Carl–Philipp Heisenberg

<!–p246 | –>doi:10.1038/nrg2548

• Abstract
• Full text
• PDF (6,227 KB)

August 2009 Volume 10 No 8

Evolutionary analysis of the dynamics of viral infectious disease

Oliver G. Pybus & Andrew Rambaut

<!–p246 | –>doi:10.1038/nrg2583

• Abstract
• Full text
• PDF (554 KB)

July 2009 Volume 10 No 7

From DNA sequence to transcriptional behaviour: a quantitative approach

Eran Segal & Jonathan Widom

<!–p246 | –>doi:10.1038/nrg2591

• Abstract
• Full text
• PDF (575 KB)

June 2009 Volume 10 No 6

Detecting gene–gene interactions that underlie human diseases

Heather J. Cordell

<!–p246 | –>doi:10.1038/nrg2579

• Abstract
• Full text
• PDF (14,371 KB)
• Supplementary information

February 2009 Volume 10 No 2

Stochastic modelling for quantitative description of heterogeneous biological systems

Darren J. Wilkinson

<!–p246 | –>doi:10.1038/nrg2509

• Abstract
• Full text
• PDF (622 KB)

MKPRF–Bustamante, Nature, 2000

A theoretical framework for extracting this information derives from considerations of the equilibrium flux of fixations and limiting probability densities of nucleotide substitutions affected by mutation, selection and random genetic drift taking place simultaneously and independently at multiple sites in a DNA sequence. In this framework known as Poisson random field (PRF), the magnitude of each cell observed in a DPRS table is an independent Poisson random variable whose expected value is given by the number of alleles sequenced from each o the species being compared.

Although each DPRS table has four parameters (_s, _a, and t) and four observations (K_a, K_s, S_a and S_s), the divergence time t is a common parameter. This implies that each gene contributes valuable information about the distribution of among genes. We therefore chose an analytical method that borrows information from all the genes to make inferences about the magnitude of selection for any individual gene. This approach greatly increases the power and accuracy of the inferences regarding selection.

A suitable framework is provided by the hierarchical Bayesian model described in Box 1 (refs 9, 10). For each species pair, we assume that the magnitude of for each gene is drawn randomly and independently from a normal distribution with mean and standard deviation . The hierarchical structure is achieved by assuming that and are themselves random variables. On the basis of this model we estimate the probability distribution of given the observed data and show that this distribution for genes in Arabidopsis is significantly different from that for genes in Drosophila.

Vaccine-autism war–Plos Biol feature

Gross, L. A broken trust: Lessons from the vaccine–autism wars. PLoS Biol 7, e1000114+ (2009). URL http://dx.doi.org/10.1371/journal.pbio.1000114

Now, more than ten years after unfounded doubts about vaccine safety first emerged, scientists and public health officials are still struggling to set the record straight. But as climate scientists know all too well, simply relating the facts of science isn’t enough. No matter that the overwhelming weight of evidence shows that climate change is real, or that vaccines don’t cause autism. When scientists find themselves just one more voice in a sea of “opinions” about a complex scientific issue, misinformation takes on a life of its own.

cis vs trans–Barkai group

Tirosh, I., Reikhav, S., Levy, A. A. & Barkai, N. A yeast hybrid provides insight into the evolution of gene expression regulation. Science 324, 659-662 (2009). URL http://dx.doi.org/10.1126/science.1169766.The message is not surprising but nevertheless quite interesting: cis-effects are mostly independent of environment while trans is mostly dependent. For the latter, another way to state the observation is that variation in multiple genes did not map to transcriptional regulators, but, in certain cases, to signal transduction genes. To me this suggests that when species adapt to different environments they change the sensory system and trans output rather than the cis regulatory systems.

Another point is that in the transcriptional regulatory system, now we can define 3 components: cis, “regulatory sense”, “sensory trans”, in which the “regulatory trans” refer to the direct TF regulating the genes, probably more about the binding specificities, while “sensory trans” indicate things upstream of that: how organisms sense the environment and the threshold at which they react as well as the extent of reaction. I’ve been seeing a lot of conservation at the regulatory sense side but the sensory side involves more stuff, definitely deserves more thinking.

Transcriptome-wide noise controls lineage choice–Chang HH, nature 2008

Chang, H. H., Hemberg, M., Barahona, M., Ingber, D. E. & Huang, S. Transcriptome-wide noise controls lineage choice in mammalian progenitor cells. Nature 453, 544-547 (2008). URL http://dx.doi.org/10.1038/nature06965

Phenotypic cell-to-cell variability within clonal populations may be a manifestation of ‘gene expression noise’^{1, 2, 3, 4, 5, 6}, or it may reflect stable phenotypic variants⁷

My thinking

the relationship between environmental or genetic upstream signals and the output (developmental, cell fate, etc) could be such that the underlying network structure already determined several stable states (multiple attractors). Noises may be unavoidable but cells might utilize the noise to switch between or choose from multiple states.

maybe what the organisms care is not the exact fate of one cell but the proportion of each type of cells, such that a probabilistic decision framework rather than a finely regulated deterministic one is employed. noise can produce order?..

really conceptual at the moment still

Ways to estimate selective coefficient–Eyre-Walker

Piganeau, G. & Eyre-Walker, A. Estimating the distribution of fitness effects from dna sequence data: Implications for the molecular clock. Proceedings of the National Academy of Sciences of the United States of America 100, 10335-10340 (2003). URL http://dx.doi.org/10.1073/pnas.1833064100

• Although our method for estimating the distribution of fitness effects is seemingly quite general, it can in practice only be applied to data sets in which there are few strongly advantageous mutations. This is because advantageous mutations decouple polymorphism and substitution: if the advantageous mutations are under directional selection, they contribute little to polymorphism, and if they are under balancing selection, they contribute little to divergence.
• Note that we implicitly assume here, and in the actual implementation of this method, that the time of divergence is much greater than the age of polymorphisms being considered, and that we can therefore ignore any contribution polymorphism makes to the apparent divergence between the two species.

Phenotypic robustness and net work centrality in yeast — Mark Siegal

• my question
  Robustness should be talked in presence of perturbations, be it genetic or environmental
  The genes found here is more like “breaking down a system by knocking out the component” and when you do it, the cells growth, development and function will gradually be affected, i.e. more chance to go wrong
• capacitors: emergent properties of complex network. From Siegal, Bergman, 2003 Nature
  “We use
  
• sources of variation in yeast cultures
  • genetic
  • growth medium
  • micro-environmental variation–neighbours
  • stochastic effect due to small number of TF
  • INPUT vs. SYSTEM-SELF
  • in this experiment, no genetic variation (var is measured within genetically identical population)
• variance depends on mean
  • lowess regression to get the residual variance (makes sense? at least a sensible thing to do)
• the 200 papameters are not all independent of each other–dimensional reduction
  • partitioning around medoids (PAM)
  • clustering to remove physically or biologically redundant phenotypes
  • 200 phenotypes -> 70 medoids
  • average top 35 (if average over 70, dilute signal; if average 2,3, then not address the question “which gene knock-out increase var broadly?”)
• how to draw significance of variance increase?
• Result
  • capacitors => more likely synthetic-lethal interactions
  • not always “essential”

Jason Gertz–Modeling synthetic yeast promoters using purely thermodynamic modelsMat

Gertz, J., Siggia, E. D. & Cohen, B. A. Analysis of combinatorial cis-regulation in synthetic and genomic promoters. Nature (2008). URLhttp://dx.doi.org/10.1038/nature07521.

• Materials
  ~2k synthetic promoters in yeast, expression measured by flow cytometry of 25,000 individual cells per promoter
• Model
  based on Buchler, Gerland and Hwa 2003, PNAS
  components include protein-DNA binding energy and protein-protein interaction, trying to predict the occupancy of RNAP as a measure of expression level.  Any chemical, enzymatic events such as chromatin modifications and so on are not included
• Results
  1. with cooperativity, one weak site + one strong site (WS) is definitely stronger than one strong site (S) and is almost as strong as two strong sites (SS)
     Q: is there a theoretical explanation for cooperativity to reduce the differences between SS and WS?
  2. I noticed from the below figure that SS seems to have a smaller variance than WS?
     
  3. Essentially this is the same idea as my potential calculation.  They claim that there thermodynamic model prediction is slightly better than simply looking for strong binding site, i.e. they can identify target promoters consisting of multiple weak sites.–
  4. Comparing two kinds of methods in predicting MIG1 regulated promoters
     1. at least one significant match to a strong motif
     2. their thermodynamic model’s output (units of repression)
     3. the first method found 359, they then take the top ranking 359 predicted promoters and overlap both sets to the 169 known MIG1 regulated promoters to find the first set by identifying strong site overlap 33 and their method overlap 41. Still they cannot explain 136-41 = 95 promoters