All speaker abstracts and talk titles will be posted as they become available.
The effects of chromatin on the dynamics of gene expression: measurements and models
In mammalian cells the state of chromatin highly dynamic and tightly liked to gene expression. It is essential to understand the rules that connect these two dynamics processes if we want to build a predictive model of gene regulation. For instance, how fast can chromatin regulators affect gene expression, how long do their effects last as epigenetic memory, and how far along the chromatin fiber do these effects spread?
In order to quantitatively address these questions, we engineered a set of mammalian cell lines that allow us to precisely control the time of recruitment and release of chromatin regulators at a fluorescent reporter gene. We recruit a set of chromatin regulators associated with various chromatin modifications: histone methylation, histone acetylation, and DNA methylation. We follow the changes in gene expression in single cells over time using time-lapse microscopy and flow cytometry. For all chromatin regulators studied, their effects can be described as stochastic transitions among three gene states: active, reversibly silent, and irreversibly silent. Mathematically, the only difference between chromatin regulators comes from the values of the stochastic transition rates among these three gene states.
Using a simple Monte Carlo simulation of chromatin modifications dynamics at a nucleosome array, we show that the stochastic nature of gene silencing and activation could arise from spreading of epigenetic modifications along the chromatin fiber. In order to experimentally test this spreading model, we use two reporter genes separated by different spacers, and target the upstream gene with different chromatin regulators. Silencing initiated at the upstream gene can rapidly spread to the downstream gene across different insulator elements and across 5000bp of spacer DNA. Epigenetic memory and reactivation of the two genes is also highly correlated at the single-cell level, suggesting that active gene expression states can also spread.
These results set the basis for a unified mathematical framework of how chromatin regulators operate to control gene expression and epigenetic memory in individual cells over time.
Understanding a cell’s molecular steering wheel
In many contexts, cells need to interpret chemically encoded spatial cues from their environment to survive or carry their physiological role. An extreme example is the chemotaxis of immune cells. These cells sense small differences in the concentration of chemical attractants and use this information to direct their movement. Furthermore, they respond to and prioritize between multiple competing attractant signals, allowing finely tuned regulation of their recruitment and dispersal during immune responses. Such regulation is critical for eliminating infections while avoiding inflammatory diseases. While it is known that the chemotaxis behaviors are driven by a family of G-protein coupled receptors, how signals are processed and integrated downstream of the receptors remains unclear. We are using quantitative live-cell imaging, including new strategies combining precise light-based control of signaling inputs and simultaneous measurement of downstream signaling outputs, to directly interrogate signal processing. Our results provide insights into how the signaling is wired to balance stable cell polarity and persistent movement with directional sensing, and how differences in signaling dynamics may mediate attractant prioritization.
Wise, benevolent, and selfish RNAs
Interactive Random Forests (iRF) and signed iRF (s-iRF)
for predictive and stable high-order interactions
Building on random forests (RFs) and random intersection trees (RITs) and through extensive biologically inspired simulations, we developed the iterative random forest algorithm (iRF) and its enhanced version signed iRF (s-iRF). iRF trains a feature-weighted ensemble of decision trees to detect predictive, stable, and high-order interactions with the same order of computational cost as the RF. s-iRF improves upon iRF by adding signs to the interactions based on new null metrics. We demonstrate the utility of iRF
and s-iRF for high order interaction discovery in genomics problems
including one that concerns enhancer activity in the early Drosophila embryo.
From Bits to Bedside: Translating Large-Scale Routine Clinical Datasets into Precision Mammography
We demonstrate how to use deep learning (DL) approaches to translate big data from routine clinical care into medical innovation that directly improves routine clinical care. Typically, large healthcare institutions have sufficient quantities of clinical data to facilitate precision medicine through a DL paradigm. However, this clinical data is hardly translated into direct clinical innovation because computer algorithms cannot readily ingest or reason over it. Using routine mammography screening data for breast cancer as an example, we first downloaded twenty years of free text pathology reports and used natural language processing to infer cancer outcomes for individual patients. We then labeled close to one million mammographic views of breast imaging with our inferred pathology outcomes. Finally, we trained convolutional neural network DL algorithms to directly predict pathology outcomes from breast imaging. With our approach, we demonstrate how to leverage DL to realize precision oncology and significantly improve the interpretation of routine screening mammography for millions of women using routine clinical big data.