Stochasticity inherent to biochemical reactions (intrinsic noise) and variability in cellular states (extrinsic noise) degrade information transmitted through signaling networks. I will discuss two recent projects from the lab that analyzed mechanism utilized by cells to mitigate noise-induced information loss. In the first project, we analyzed the ability of temporal signal modulation--that is, dynamics--to reduce noise-induced information loss. In the extracellular signal-regulated kinase (ERK), calcium (Ca(2+)), and nuclear factor kappa-B (NF-κB) pathways, response dynamics resulted in significantly greater information transmission capacities compared to nondynamic responses. Theoretical analysis demonstrated that signaling dynamics has a key role in overcoming extrinsic noise. Experimental measurements of information transmission in the ERK network under varying signal-to-noise levels confirmed our predictions and showed that signaling dynamics mitigate, and can potentially eliminate, extrinsic noise-induced information loss. By curbing the information-degrading effects of cell-to-cell variability, dynamic responses substantially increase the accuracy of biochemical signaling networks. In the second project, we investigated how cells utilize cell to cell communication to reduce the effects of noise using epidermal growth factor receptor (EGFR) transactivation that results from extracellular ATP, a key damage signal critical for wound response signaling. We show that ectodomain shedding results in paracrine communication that reduces the variability of pathways downstream of EGFR. Analysis of cellular response fidelity as a function of the paracrine communication distance showed that the experimentally measured communication distance maximizes the reliability of wound response signals. Our results demonstrate that local paracrine communication can be used to filter out response noise through local averaging and that this noise filtering can be optimized by regulating paracrine communication distance.

About the Speaker

Faculty Host: Robin Lee

One of the key questions in my lab deals with the proper identification of causal links between molecular processes in complex pathways. We define complexity as the product of high nonlinearity and high redundancy between processes. Intrinsic to such systems are adaptive responses to perturbations. Therefore, conventional molecular and genetic intervention studies often fail in providing information of the function of a targeted systems component. In this overview talk I will take cell protrusion as a prime example of a cell functional outcome of a pathway system with such properties. I will introduce the mathematical, computational, and experimental concepts of image fluctuation analysis as a method for accurate delineation of the functional hierarchy between molecular and mechanical processes driving cell morphogenic events. I will motivate this fundamental systems biological problem with experiments that highlight the mechanisms of action of several oncogenes in driving metastatic cell migration.

It has long been proposed that turgor pressure plays an essential role during bacterial growth by driving mechanical expansion of the cell wall. This hypothesis is based on analogy to plant cells, for which this mechanism has been established, and on experiments in which the growth rate of bacterial cultures was observed to decrease as the osmolarity of the growth medium was increased. To distinguish the effect of turgor pressure from pressure-independent effects that osmolarity might have on cell growth, we monitored the elongation of single Escherichia coli cells while rapidly changing the osmolarity of their media. By plasmolyzing cells, we found that cell-wall elastic strain did not scale with growth rate, suggesting that pressure does not drive cell-wall expansion. Furthermore, in response to hyper- and hypoosmotic shock, E. coli cells resumed their pre-shock growth rate and relaxed to their steady-state rate after several minutes, demonstrating that osmolarity modulates growth rate slowly, independently of pressure. Oscillatory hyperosmotic shock revealed that while plasmolysis slowed cell elongation, the cells nevertheless “stored” growth such that once turgor was re-established the cells elongated to the length that they would have attained had they never been plasmolyzed. In contrast, Bacillus subtilis cells exhibit highly regular growth oscillations in response to hypoosmotic shock that are dependent on peptidoglycan synthesis. The period of these oscillations scales linearly with the magnitude of the shock. By applying a simple mathematical theory to these data, we show that growth oscillations are initiated by mechanical-strain-induced growth arrest. This demonstrates that B. subtilis has developed an elegant system by which turgor pressure both up- and down-regulates the final steps of cell growth.

About the Speaker

Juniors & Seniors!

Interested in getting a graduate degree from CMU?
Are you interested in curing biomedical disorders?
Do you want to analyze large and complex data sets?
Do you love solving difficult real-world problems?

Program Facts

  • Interdisciplinary program with the Department of Biological Sciences and the Computational BIology Department
  • Opportunity for hands on research
  • Wide array of electives
  • Focus on professional development and networking

Meet the program advisors and find out how to finish in one year.


Genome regions vary widely in their evolutionary rates. Not only may mutation rates differ among regions, but so too the effects of forces acting on mutations such as selection, recombination and genetic drift. One of our goals is to disentangle the contributions of these processes to evolutionary rates and patterns of gene organization, which would both improve the legibility of genomes and deepen our understanding of problems caused by deleterious mutations such as species extinction or somatic disease states like cancer and dementia. The remarkable lack of evidence for the nature of mutation, especially in prokaryotes, motivated a series of mutation-accumulation (MA) experiments using bacteria with multiple chromosomes, in which secondary, smaller chromosomes evolve much faster than the primary chromosome. MA experiments allow mutations to accumulate in many replicate lines in the near absence of selection, and whole-genome sequencing of evolved lines allows us to capture mutational distributions and any inherent biases.

Here, we describe genome-wide views of the molecular mutation spectrum in three distinct bacterial species (Burkholderia cenocepacia, Vibrio cholerae, and Vibrio fischeri) whose genomes all harbor multiple chromosomes but vary substantially in %GC content. All three species have low mutation rates with insertion-deletion mutations biased towards deletions. While the direction of mutation bias is consistent with the realized GC-content of all three organisms, realized GC-content is always higher than predicted by mutation pressure alone. We also observed variation in both the rates and spectra of mutations among chromosomes, and a significant elevation of G:C>T:A transversions in late-replicating regions that is consistent with greater oxidative damage later in the cell cycle. Most intriguing, mutation rates exhibit a wave-like pattern correlated with replication timing, again suggesting that the cellular state undergoes fluctuations that influence genome stability. Collectively, these findings support theory that genome architectures and content are shaped by systematic differences in both the origin and effects of mutations.

About the Speaker.

A mechanistic understanding of biological signaling networks is typically hindered by the complexity associated with measuring all relevant parameters at multiple time points. Mathematical representations are central to understanding these complex biological signaling processes but, as signaling networks increase in size and complexity, it becomes increasingly challenging to manage biological knowledge in mathematical form and link these models to experimental data. Here we show a novel paradigm whereby we leverage the power of rule-based languages to encode biological models of signaling processes as Python programs thorough our PySB models-as-programs framework. We show  how our modeling approach can be used to explore network topologies and guide experiments to understand how cells commit to either apoptosis or necroptosis forms of programmed cell death. We also demonstrate how our approach can be used to explore parameter uncertainties in complex kinetic models and identify the key parameters that drive dynamics in COX-2 kinetics. We showcase how PySB models, as Python programs, can leverage tools and practices from the open-source software community, substantially advancing our ability to collaborate, disseminate knowledge, and manage the testing of biochemical hypotheses. Our modeling paradigm  guides experiments and makes the theory-and-experiments cycle easily accessible to users.

About the Speaker


Chemical shifts represent the most ubiquitous and accurately measured NMR parameters of proteins and their knowledge is a prerequisite for the interpretation of most NMR experiments. To date, chemical shifts of thousands of proteins have been deposited in the BioMagResDataBank. On the other hand, chemical shifts belong to the most difficult NMR parameters in terms of their interpretation because of their highly sensitive dependence on their local chemical environment. In addition, the experimental chemical shift of a given nucleus reflects the Boltzmann-weighted average of the 'instantaneous' chemical shifts of a large number of conformational sub-states that interconvert on the millisecond timescale or faster, which are hard to capture. Chemical shift information has found widespread use in protein NMR, such as for the assessment of the propensity of protein segments to adopt various types of secondary structure, the determination and refinement of 3D protein structures, and the extraction of site-specific order parameters as measures of local dynamics. In this talk, I will discuss some of our recent work from two sides of the protein chemical shift problem: (1) the prediction of chemical shifts through in silico protein ensembles and (2) the improvement of molecular dynamics simulations through their protein force fields using experimental chemical shifts of native proteins as a guide.

Advances in high-throughput technology over the last decade allowed us to capture biologically meaningful data on a scale not possible before. Carefully designed experimental protocols now allow scientists to capture protein-protein interactions (PPI), dependencies between genes, their products, and metabolites for tens of thousands of targets at a time. At the same time, rapid evolution of sequencing technologies made whole genome and transcriptome analyses possible without compromising their accuracy. However, large amounts of data generated by these experiments make data storage and transmission, as well as analyses and visualization of the data, difficult.

In this dissertation, we explore solutions to the data storage problem and suggest approaches to sequence and sequence alignment compression. Specifically, we consider a general problem of identifying information shared across multiple related datasets and its applications to compression of whole file collections. We formulate the problem of encoding the data given such shared information and provide an algorithm for doing so optimally. Further, we develop a functional compression scheme for sequence alignments that, while outperforming state of the art tools in terms of compression performance, also allows for random access to data, fast computations on compressed sequence, and modular data downloads. Additionally, we investigate the problem of visualizing large multi-dimensional datasets where we are interested in evaluating the robustness of annotations assigned to the data points. We develop novel ways of representing such data that helps to assess the diversity of annotations and find groups of the most consistent annotations. As an example, we study the protein function imputation for a PPI network of an A. thaliana plant. We continue by investigating challenges of visualizing such complex data when data is changing over time and forms a graph. We propose a rich visual system that focuses on exposing the graph’s local structure while allowing the system’s state at multiple timepoints to influence the current view. We further consider chromosome conformation data that, for the first time, captures the 3D conformation of nuclear DNA. We apply the strategies for identifying robust classification subgroups to the problem of inferring densely packed contiguous topological domains. Our approach provides a scalable way to find such topological domains and to walk the space of near-optimal structures giving the first opportunity to quantify DNA’s hierarchical spatial structure. The visualization and algorithmic approaches we developed are scalable and can be applied outside of the biological data domain: for example, the exploration of annotations and their robustness is of importance for any set of predictions.

Thesis Committee:
Carl Kingsford (Advisor)
Russell Schwartz
Takis Benos (University of Pittsburgh)
Liz Marai (University of Pittsburgh)

Development of high-throughput monitoring technologies enables interrogation of cancer cells at various levels of cellular activity. Various public efforts such as The Cancer Genome Atlas (TCGA) generate disparate omic data for large patient cohorts, and these heterogeneous data sources provide the opportunity to gain insights into the molecular changes that drive cancer pathogenesis and progression. However, these insights are limited by the vast search space and as a result low statistical power to make new discoveries. In this talk, I describe methods for extracting informative features from disparate omic data, with a view to gaining mechanistic insights into the relationship between molecular changes at different levels of cellular activity. The results described herein show that one can integrate disparate omic data using molecular interaction networks, providing insights beyond what can be gleaned by analyzing different types of molecular data in isolation.

Faculty Host: Ziv Bar-Joseph

Next generation sequencing platforms have revolutionized the field of biology in the past decade. However, inherent noise processes at the biochemistry level and signal acquisition stage warrant the use of corrective algorithms to “clean” the data and/or “make sense” of the data. Given the large amounts of sequenced data generated by present sequencing platforms for a plethora of downstream applications, such algorithms need to be both accurate and scalable

In this talk, I will highlight my work on two such problems occurring in the field of bioinformatics/computational biology. First, for the problem of basecalling for sequencing-by-synthesis (Illumina) platforms, I describe novel computationally tractable statistical models and signal processing schemes that are fast and have lower error rates than other state-of-the-art basecallers. Extensions to a soft information exchange setup to do joint basecalling and SNP calling are also explored. Next, I describe two novel single individual haplotyping inference schemes using an (optimal) branch and bound framework and (scalable) low rank semidefinite programming ideas for diploid and polyploid species. Ongoing work in this regard include detection of unknown ploidy and joint genotyping-haplotyping.


Subscribe to CBD