Publications

We describe a computational framework for simulating suspensions of rigid particles in Newtonian Stokes flow. One central building block is a collision-resolution algorithm that overcomes the numerical constraints arising from particle collisions. This algorithm extends the well-known complementarity method for non-smooth multi-body dynamics to resolve collisions in dense rigid body suspensions. This approach formulates the collision resolution problem as a linear complementarity problem with geometric `non-overlapping' constraints imposed at each timestep. It is then reformulated as a constrained quadratic programming problem and the Barzilai-Borwein projected gradient descent method is applied for its solution. This framework is designed to be applicable for any convex particle shape, e.g., spheres and spherocylinders, and applicable to any Stokes mobility solver, including the Rotne-Prager-Yamakawa approximation, Stokesian Dynamics, and PDE solvers (e.g., boundary integral and immersed boundary methods). In particular, this method imposes Newton's Third Law and records the entire contact network. Further, we describe a fast, parallel, and spectrally-accurate boundary integral method tailored for spherical particles, capable of resolving lubrication effects. We show weak and strong parallel scalings up to 8×104 particles with approximately 4×107 degrees of freedom on 1792 cores. We demonstrate the versatility of this framework with several examples, including sedimentation of particle clusters, and active matter systems composed of ensembles of particles driven to rotate.

The translational power of human microbiome studies is limited by high interindividual variation. We describe a dimensionality reduction tool, compositional tensor factorization (CTF), that incorporates information from the same host across multiple samples to reveal patterns driving differences in microbial composition across phenotypes. CTF identifies robust patterns in sparse compositional datasets, allowing for the detection of microbial changes associated with specific phenotypes that are reproducible across datasets.

Fibrin is the three-dimensional mechanical scaffold of protective blood clots that stop bleeding and pathological thrombi that obstruct blood vessels. Fibrin must be mechanically tough to withstand rupture, after which life-threatening pieces (thrombotic emboli) are carried downstream by blood flow. Despite multiple studies on fibrin viscoelasticity, mechanisms of fibrin rupture remain unknown. Here, we examined mechanically and structurally the strain-driven rupture of human blood plasma–derived fibrin clots where clotting was triggered with tissue factor. Toughness, i.e., resistance to rupture, quantified by the critical energy release rate (a measure of the propensity for clot embolization) of physiologically relevant fibrin gels was determined to be 7.6 ± 0.45 J/m2. Finite element (FE) simulations using fibrin material models that account for forced protein unfolding independently supported this measured toughness and showed that breaking of fibers ahead the crack at a critical stretch is the mechanism of rupture of blood clots, including thrombotic embolization.

Tissue morphogenesis relies on repeated use of dynamic behaviors at the levels of intracellular structures, individual cells, and cell groups. Rapidly accumulating live imaging datasets make it increasingly important to formalize and automate the task of mapping recurrent dynamic behaviors (motifs), as it is done in speech recognition and other data mining applications. Here, we present a “template-based search” approach for accurate mapping of sub- to multi-cellular morphogenetic motifs using a time series data mining framework. We formulated the task of motif mapping as a subsequence matching problem and solved it using dynamic time warping, while relying on high throughput graph-theoretic algorithms for efficient exploration of the search space. This formulation allows our algorithm to accurately identify the complete duration of each instance and automatically label different stages throughout its progress, such as cell cycle phases during cell division. To illustrate our approach, we mapped cell intercalations during germband extension in the early Drosophila embryo. Our framework enabled statistical analysis of intercalary cell behaviors in wild-type and mutant embryos, comparison of temporal dynamics in contracting and growing junctions in different genotypes, and the identification of a novel mode of iterative cell intercalation. Our formulation of tissue morphogenesis using time series opens new avenues for systematic decomposition of tissue morphogenesis.

Microglial cells play a key role in brain homeostasis from development to adulthood. Here we show the involvement of a site-specific phosphorylation of Presenilin 1 (PS1) in microglial development. Profiles of microglia-specific transcripts in different temporal stages of development, combined with multiple systematic transcriptomic analysis and quantitative determination of microglia progenitors, indicate that the phosphorylation of PS1 at serine 367 is involved in the temporal dynamics of microglial development, specifically in the developing brain rudiment during embryonic microgliogenesis. We constructed a developing brain-specific microglial network to identify transcription factors linked to PS1 during development. Our data showed that PS1 functional connections appear through interaction hubs at Pu.1, Irf8 and Rela-p65 transcription factors. Finally, we showed that the total number of microglia progenitors was markedly reduced in the developing brain rudiment of embryos lacking PS1 phosphorylation compared to WT. Our work identifies a novel role for PS1 in microglial development.

Tissue morphogenesis relies on repeated use of dynamic behaviors at the levels of intracellular structures, individual cells, and cell groups. Rapidly accumulating live imaging datasets make it increasingly important to formalize and automate the task of mapping recurrent dynamic behaviors (motifs), as it is done in speech recognition and other data mining applications. Here, we present a “template-based search” approach for accurate mapping of sub- to multi-cellular morphogenetic motifs using a time series data mining framework. We formulated the task of motif mapping as a subsequence matching problem and solved it using dynamic time warping, while relying on high throughput graph-theoretic algorithms for efficient exploration of the search space. This formulation allows our algorithm to accurately identify the complete duration of each instance and automatically label different stages throughout its progress, such as cell cycle phases during cell division. To illustrate our approach, we mapped cell intercalations during germband extension in the early Drosophila embryo. Our framework enabled statistical analysis of intercalary cell behaviors in wild-type and mutant embryos, comparison of temporal dynamics in contracting and growing junctions in different genotypes, and the identification of a novel mode of iterative cell intercalation. Our formulation of tissue morphogenesis using time series opens new avenues for systematic decomposition of tissue morphogenesis.

Convolutional neural networks (CNNs) have become a standard for analysis of biological sequences. Tuning of network architectures is essential for CNN’s performance, yet it requires substantial knowledge of machine learning and commitment of time and effort. This process thus imposes a major barrier to broad and effective application of modern deep learning in genomics. Here, we present AMBER, a fully automated framework to efficiently design and apply CNNs for genomic sequences. AMBER designs optimal models for user-specified biological questions through the state-of-the-art Neural Architecture Search (NAS). We applied AMBER to the task of modelling genomic regulatory features and demonstrated that the predictions of the AMBER-designed model are significantly more accurate than the equivalent baseline non-NAS models and match or even exceed published expert-designed models. Interpretation of AMBER architecture search revealed its design principles of utilizing the full space of computational operations for accurately modelling genomic sequences. Furthermore, we illustrated the use of AMBER to accurately discover functional genomic variants in allele-specific binding and disease heritability enrichment. AMBER provides an efficient automated method for designing accurate deep learning models in genomics.

Vancomycin-resistant Enterococcus faecium (VREfm) is an emerging antibiotic-resistant pathogen. Strain-level investigations are beginning to reveal the molecular mechanisms used by VREfm to colonize regions of the human bowel. However, the role of commensal bacteria during VREfm colonization, in particular following antibiotic treatment, remains largely unknown. We employed amplicon 16S rRNA gene sequencing and metabolomics in a murine model system to try and investigate functional roles of the gut microbiome during VREfm colonization. First-order taxonomic shifts between Bacteroidetes and Tenericutes within the gut microbial community composition were detected both in response to pretreatment using ceftriaxone and to subsequent VREfm challenge. Using neural networking approaches to find cooccurrence profiles of bacteria and metabolites, we detected key metabolome features associated with butyric acid during and after VREfm colonization. These metabolite features were associated with Bacteroides, indicative of a transition toward a preantibiotic naive microbiome. This study shows the impacts of antibiotics on the gut ecosystem and the progression of the microbiome in response to colonization with VREfm. Our results offer insights toward identifying potential nonantibiotic alternatives to eliminate VREfm through metabolic reengineering to preferentially select for Bacteroides.

IL-4 activates macrophages to adopt distinct phenotypes associated with clearance of helminth infections and tissue repair, but the phenotype depends on the cellular lineage of these macrophages. The molecular basis of chromatin remodeling in response to IL-4 stimulation in tissue-resident and monocyte-derived macrophages is not understood. In this study, we find that IL-4 activation of different lineages of peritoneal macrophages in mice is accompanied by lineage-specific chromatin remodeling in regions enriched with binding motifs of the pioneer transcription factor PU.1. PU.1 motif is similarly associated with both tissue-resident and monocyte-derived IL-4-induced accessible regions but has different lineage-specific DNA shape features and predicted cofactors. Mutation studies based on natural genetic variation between C57BL/6 and BALB/c mouse strains indicate that accessibility of these IL-4-induced regions can be regulated through differences in DNA shape without direct disruption of PU.1 motifs. We propose a model whereby DNA shape features of stimulation-dependent genomic elements contribute to differences in the accessible chromatin landscape of alternatively activated macrophages on different genetic backgrounds that may contribute to phenotypic variations in immune responses.

Clustered Regularly Interspace Short Palindromic Repeats (CRISPR)-Cas9 genome editing methods provide the tools necessary to examine phenotypic impacts of targeted perturbations in high-throughput screens. While these technologies have the potential to reveal functional elements with direct therapeutic applications, statistical techniques to analyze noncoding screen data remain limited. We present CRISPR-Decryptr, a computational tool for the analysis of CRISPR noncoding screens. Our method leverages experimental design: accounting for multiple conditions, controls, and replicates to infer the regulatory landscape of noncoding genomic regions. We validate our method on a variety of mutagenesis, CRISPR activation, and CRISPR interference screens, extracting new insights from previously published data.