Intracellular information transfer: understanding bit by bit
Host: S. Jamal Rahi
Signal transduction pathways have evolved to enable sensing of changes
in extracellular and intracellular conditions and triggering
appropriate cell responses to such changes. Our understanding of the
information transfer capabilities of signaling networks is still very
limited, particular of those in eukaryotic cells and organisms. In
this talk, I will illustrate how an understanding of the signaling
network structure and function can be achieved through progressive
iterations between model and experiment and how this analysis can be
further extended by incorporating new languages and concepts,
particularly suitable to information processing. I will use the
example of the tumor necrosis factor-stimulated network to illustrate
this analysis and to discuss the limitations and capabilities of
signaling networks more generally.
New approaches to studying the growth and size regulation of mammalian cells
Host: S. Jamal Rahi
The study of cell growth has been limited primarily by the lack of accurate enough means of measuring the growth of cells as they traverse the cell cycle. There are several theoretical models of growth that have been impossible to evaluate because the methods for measuring growth have been too inaccurate to distinguish among them. In particular, if cells grow proportional to their mass, which of course doubles each cell cycle, then it is likely that the variation in cell size in a population would increase without limit. This is simply because cell division is rarely completely symmetric, producing smaller cells that would grow slower and larger cells that would grow faster. On the other hand, if cells added equal mass per unit time this undesirable outcome could be avoided. There are ideas that size control may not exist but simply be driven by exogenous and independent controls of cell cycle and growth, size being simply a resultant of these explicity controls. Yet the very strict size regulation of different cell types, suggests that cell size is an evolutionary optimum for different functions and hence, cells should have a homeostatic mechanism for maintaining cell size. There are other speculations that cells grow to a defined size and then divide, making cell division a slave to cell growth. The opposite is also possible that passage through the cell cycle feeds back on cell growth. To approach these questions we have developed two new analytical techniques of exquisite sensitivity. In collaboration with Scott Manalis at MIT, we used his suspended microchannel resonator to measure cell mass to 0.01% and to do that for as many as 8 generations without causing any known harm to the cells. This technique pointed to a sharp transition of growth at the G1/S transition. It also shows that a size threshold does not exist in a mammalian cell line but instead there is convergence of cell growth rates at G1/S. Another technique which Ran Kafri, a postdoc in my lab and Galit Lahav’s lab developed used a static population based approach to derive very sensitive kinetic features based on the ergodic assumption of steady state growth. This method opens up many new measurements not possible in growing individual cells; here temporal resolution and sensitivity is increased markedly as cell numbers exceed a million. This method also described a period at the feedback on growth rate at the G1/S transition. These new measurements suggest that there is a sizing mechanism in mammalian cells that reduces variation in the cell cycle by affecting growth rate and size dependence of growth rate . Such a mechanism is liked to be tuned and respond differently in different cell types and under different conditions.
Characterizing transcriptomes from high throughput sequencing data: from yeast to mammals
Host: E. Siggia
Experimentally defining the complete transcriptome of eukaryotic organisms has traditionally been a challenging task, but advances in sequencing RNA (RNAseq) offer new and powerful approaches to the study of transcriptomes. Most studies have used RNAseq to quantify the expression levels of known genes, identify splice isoforms and refine gene boundaries. However, many studies depend on an existing annotation or sequenced genomes, limiting the ability of discovering novel transcripts and studying diverse organisms.
I will present a series of studies on the development of technologies and tools for RNAseq analysis and their application in organisms ranging from yeast to mouse. I will focus on different approaches I have developed for transcriptome reconstruction, from mapping-first ones that rely only on an available genome sequence, to Trinity a method for de novo assembly of full-length transcripts without requiring a sequenced genome, but with a sensitivity similar to methods that rely on genome alignments. In addition, I will describe systematic approaches to assess the quality of RNA-Seq experiments for annotation and expression quantification, and how we used them in a comparative study on library construction methods for strand specific RNAseq.
Finally, I will show how these approaches scale to organisms from yeasts to vertebrates, helping in genome annotation of newly discovered organisms from the Schizosaccharomyces clade, the identification of extensive regulated long antisense transcripts that are conserved across yeast species, transcriptome analysis in the Bemisia tabaci whitefly, for which the genome sequence is not available, and for the discovery of alternatively spliced isoform in mouse.
Over a lifetime the brain performs a large number of individual cognitive acts, most having some dependence on past experience. It is difficult to reconcile such capabilities, even in principle, with the known resource constraints on cortex, such as low connectivity and low average synaptic strength. Here we describe model neural circuits and associated algorithms that respect the brain's most basic resource constraints and support the execution of high numbers of cognitive actions. Our circuits simultaneously support a suite of four basic kinds of task that each requires some circuit modification: hierarchical memory formation, association, supervised memorization, and inductive learning of threshold functions. The capacity of our circuits is established via experiments (joint with Vitaly Feldman) in which sequences of thousands of such acts are simulated by computer and the circuits created tested for subsequent efficacy. Our underlying theory is apparently the only biologically plausible systems-level theory of learning and memory in cortex for which such a demonstration has been performed. The theory also makes a new hypothesis about the computational role of the hippocampus.
