Publications

Organisms from all domains of life use gene regulation networks to control cell growth, identity, function, and responses to environmental challenges. Although accurate global regulatory models would provide critical evolutionary and functional insights, they remain incomplete, even for the best studied organisms. Efforts to build comprehensive networks are confounded by challenges including network scale, degree of connectivity, complexity of organism–environment interactions, and difficulty of estimating the activity of regulatory factors. Taking advantage of the large number of known regulatory interactions in Bacillus subtilis and two transcriptomics datasets (including one with 38 separate experiments collected specifically for this study), we use a new combination of network component analysis and model selection to simultaneously estimate transcription factor activities and learn a substantially expanded transcriptional regulatory network for this bacterium. In total, we predict 2,258 novel regulatory interactions and recall 74% of the previously known interactions. We obtained experimental support for 391 (out of 635 evaluated) novel regulatory edges (62% accuracy), thus significantly increasing our understanding of various cell processes, such as spore formation.

Positive-Unlabeled (PU) learning scenarios are a class of semi-supervised learning where only a fraction of the data is labeled, and all available labels are positive. The goal is to assign correct (positive and negative) labels to as much data as possible. Several important learning problems fall into the PU-learning domain, as in many cases the cost and feasibility of obtaining negative examples is prohibitive. In addition to the positive-negative disparity the overall cost of labeling these datasets typically leads to situations where the number of unlabeled examples greatly outnumbers the labeled. Accordingly, we perform several experiments, on both synthetic and real-world datasets, examining the performance of state of the art PU-learning algorithms when there is significant bias in the labeling process. We propose novel PU algorithms and demonstrate that they outperform the current state of the art on a variety of benchmarks. Lastly, we present a methodology for removing the costly parameter-tuning step in a popular PU algorithm.

Circuit diagrams of brains are generally reported only as absolute or consensus networks; these diagrams fail to identify the accuracy of connections, however, for which multiple circuits of the same neurons must be documented. For this reason, the modular composition of the Drosophila visual system, with many identified neuron classes, is ideal. Using EM, we identified synaptic connections in the fly’s second visual relay neuropil, or medulla, in the 20 neuron classes in a so-called “core connectome,” those neurons present in seven neighboring columns. These connections identify circuits for motion. Their error rates for wiring reveal that <1% of contacts overall are not part of a consensus circuit but incorporate errors of either omission or commission. Autapses are occasionally seen.

RORγt+ Th17 cells are important for mucosal defenses but also contribute to autoimmune disease. They accumulate in the intestine in response to microbiota and produce IL-17 cytokines. Segmented filamentous bacteria (SFB) are Th17-inducing commensals that potentiate autoimmunity in mice. RORγt+ T cells were induced in mesenteric lymph nodes early after SFB colonization and distributed across different segments of the gastrointestinal tract. However, robust IL-17A production was restricted to the ileum, where SFB makes direct contact with the epithelium and induces serum amyloid A proteins 1 and 2 (SAA1/2), which promote local IL-17A expression in RORγt+ T cells. We identified an SFB-dependent role of type 3 innate lymphoid cells (ILC3), which secreted IL-22 that induced epithelial SAA production in a Stat3-dependent manner. This highlights the critical role of tissue microenvironment in activating effector functions of committed Th17 cells, which may have important implications for how these cells contribute to inflammatory disease.

The cohesin complex (consisting of Rad21, Smc1a, Smc3, and Stag2 proteins) is critically important for proper sister chromatid separation during mitosis. Mutations in the cohesin complex were recently identified in a variety of human malignancies including acute myeloid leukemia (AML). To address the potential tumor-suppressive function of cohesin in vivo, we generated a series of shRNA mouse models in which endogenous cohesin can be silenced inducibly. Notably, silencing of cohesin complex members did not have a deleterious effect on cell viability. Furthermore, knockdown of cohesin led to gain of replating capacity of mouse hematopoietic progenitor cells. However, cohesin silencing in vivo rapidly altered stem cells homeostasis and myelopoiesis. Likewise, we found widespread changes in chromatin accessibility and expression of genes involved in myelomonocytic maturation and differentiation. Finally, aged cohesin knockdown mice developed a clinical picture closely resembling myeloproliferative disorders/neoplasms (MPNs), including varying degrees of extramedullary hematopoiesis (myeloid metaplasia) and splenomegaly. Our results represent the first successful demonstration of a tumor suppressor function for the cohesin complex, while also confirming that cohesin mutations occur as an early event in leukemogenesis, facilitating the potential development of a myeloid malignancy.

Many important processes in the cell are mediated by stiff microtubule polymers and the active motor proteins moving on them. This includes the transport of subcellular structures (nuclei, chromosomes, organelles) and the self-assembly and positioning of the mitotic spindle. Little is understood of these processes, but they present fascinating problems in fluid-structure interactions. Microtubules and motor proteins are also the building blocks of new biosynthetic active suspensions driven by motor-protein activity. These reduced systems can be probed—and modeled—more easily than can the fully biological ones and demonstrate their own aspects of self-assembly and complex dynamics. I review recent work modeling such systems as fluid-structure interaction problems and as multiscale complex fluids.

We reconstructed the synaptic circuits of seven columns in the second neuropil or medulla behind the fly’s compound eye. These neurons embody some of the most stereotyped circuits in one of the most miniaturized of animal brains. The reconstructions allow us, for the first time to our knowledge, to study variations between circuits in the medulla’s neighboring columns. This variation in the number of synapses and the types of their synaptic partners has previously been little addressed because methods that visualize multiple circuits have not resolved detailed connections, and existing connectomic studies, which can see such connections, have not so far examined multiple reconstructions of the same circuit. Here, we address the omission by comparing the circuits common to all seven columns to assess variation in their connection strengths and the resultant rates of several different and distinct types of connection error. Error rates reveal that, overall, <1% of contacts are not part of a consensus circuit, and we classify those contacts that supplement (E+) or are missing from it (E−). Autapses, in which the same cell is both presynaptic and postsynaptic at the same synapse, are occasionally seen;two cells in particular, Dm9 and Mi1, form ≥20-fold more autapses than do other neurons. These results delimit the accuracy of developmental events that establish and normally maintain synaptic circuits with such precision, and thereby address the operation of such circuits. They also establish a precedent for error rates that will be required in the new science of connectomics.

Many functionally important interactions between genes and proteins involved in immunological diseases and processes are unknown. The exponential growth in public high-throughput data offers an opportunity to expand this knowledge. To unlock human-immunology-relevant insight contained in the global biomedical research effort, including all public high-throughput datasets, we performed immunological-pathway-focused Bayesian integration of a comprehensive, heterogeneous compendium comprising 38,088 genome-scale experiments. The distillation of this knowledge into immunological networks of functional relationships between molecular entities (ImmuNet), and tools to mine this resource, are accessible to the public at http://immunet.princeton.edu. The predictive capacity of ImmuNet, established by rigorous statistical validation, is easily accessed by experimentalists to generate data-driven hypotheses. We demonstrate the power of this approach through the identification of unique host-virus interaction responses, and we show how ImmuNet complements genetic studies by predicting disease-associated genes. ImmuNet should be widely beneficial for investigating the mechanisms of the human immune system and immunological diseases.

Identifying functional effects of noncoding variants is a major challenge in human genetics. To predict the noncoding-variant effects de novo from sequence, we developed a deep learning–based algorithmic framework, DeepSEA (http://deepsea.princeton.edu/), that directly learns a regulatory sequence code from large-scale chromatin-profiling data, enabling prediction of chromatin effects of sequence alterations with single-nucleotide sensitivity. We further used this capability to improve prioritization of functional variants including expression quantitative trait loci (eQTLs) and disease-associated variants.

We estimated ideological preferences of 3.8 million Twitter users and, using a data set of nearly 150 million tweets concerning 12 political and nonpolitical issues, explored whether online communication resembles an “echo chamber” (as a result of selective exposure and ideological segregation) or a “national conversation.” We observed that information was exchanged primarily among individuals with similar ideological preferences in the case of political issues (e.g., 2012 presidential election, 2013 government shutdown) but not many other current events (e.g., 2013 Boston Marathon bombing, 2014 Super Bowl). Discussion of the Newtown shootings in 2012 reflected a dynamic process, beginning as a national conversation before transforming into a polarized exchange. With respect to both political and nonpolitical issues, liberals were more likely than conservatives to engage in cross-ideological dissemination; this is an important asymmetry with respect to the structure of communication that is consistent with psychological theory and research bearing on ideological differences in epistemic, existential, and relational motivation. Overall, we conclude that previous work may have overestimated the degree of ideological segregation in social-media usage.