Deep Neural Networks for Predicting Single-Cell Responses and Probability Landscapes.

Engineering biology relies on the accurate prediction of cell responses. However, making these predictions is challenging for a variety of reasons, including the stochasticity of biochemical reactions, variability between cells, and incomplete information about underlying biological processes. Machine learning methods, which can model diverse input-output relationships without requiring a priori mechanistic knowledge, are an ideal tool for this task. For example, such approaches can be used to predict gene expression dynamics given time-series data of past expression history. To explore this application, we computationally simulated single-cell responses, incorporating different sources of noise and alternative genetic circuit designs. We showed that deep neural networks trained on these simulated data were able to correctly infer the underlying dynamics of a cell response even in the presence of measurement noise and stochasticity in the biochemical reactions. The training set size and the amount of past data provided as inputs both affected prediction quality, with cascaded genetic circuits that introduce delays requiring more past data. We also tested prediction performance on a bistable auto-activation circuit, finding that our initial method for predicting a single trajectory was fundamentally ill-suited for multimodal dynamics. To address this, we updated the network architecture to predict the entire distribution of future states, showing it could accurately predict bimodal expression distributions. Overall, these methods can be readily applied to the diverse prediction tasks necessary to predict and control a variety of biological circuits, a key aspect of many synthetic biology applications.

The computational capabilities of many-to-many protein interaction networks.

Many biological circuits comprise sets of protein variants that interact with one another in a many-to-many, or promiscuous, fashion. These architectures can provide powerful computational capabilities that are especially critical in multicellular organisms. Understanding the principles of biochemical computations in these circuits could allow more precise control of cellular behaviors. However, these systems are inherently difficult to analyze, due to their large number of interacting molecular components, partial redundancies, and cell context dependence. Here, we discuss recent experimental and theoretical advances that are beginning to reveal how promiscuous circuits compute, what roles those computations play in natural biological contexts, and how promiscuous architectures can be applied for the design of synthetic multicellular behaviors.

The context-dependent, combinatorial logic of BMP signaling.

Cell-cell communication systems typically comprise families of ligand and receptor variants that function together in combinations. Pathway activation depends on the complex way in which ligands are presented extracellularly and receptors are expressed by the signal-receiving cell. To understand the combinatorial logic of such a system, we systematically measured pairwise bone morphogenetic protein (BMP) ligand interactions in cells with varying receptor expression. Ligands could be classified into equivalence groups based on their profile of positive and negative synergies with other ligands. These groups varied with receptor expression, explaining how ligands can functionally replace each other in one context but not another. Context-dependent combinatorial interactions could be explained by a biochemical model based on the competitive formation of alternative signaling complexes with distinct activities. Together, these results provide insights into the roles of BMP combinations in developmental and therapeutic contexts and establish a framework for analyzing other combinatorial, context-dependent signaling systems.

Ligand-receptor promiscuity enables cellular addressing.

In multicellular organisms, secreted ligands selectively activate, or "address," specific target cell populations to control cell fate decision-making and other processes. Key cell-cell communication pathways use multiple promiscuously interacting ligands and receptors, provoking the question of how addressing specificity can emerge from molecular promiscuity. To investigate this issue, we developed a general mathematical modeling framework based on the bone morphogenetic protein (BMP) pathway architecture. We find that promiscuously interacting ligand-receptor systems allow a small number of ligands, acting in combinations, to address a larger number of individual cell types, defined by their receptor expression profiles. Promiscuous systems outperform seemingly more specific one-to-one signaling architectures in addressing capability. Combinatorial addressing extends to groups of cell types, is robust to receptor expression noise, grows more powerful with increases in the number of receptor variants, and is maximized by specific biochemical parameter relationships. Together, these results identify design principles governing cellular addressing by ligand combinations.

Combinatorial Signal Perception in the BMP Pathway.

The bone morphogenetic protein (BMP) signaling pathway comprises multiple ligands and receptors that interact promiscuously with one another and typically appear in combinations. This feature is often explained in terms of redundancy and regulatory flexibility, but it has remained unclear what signal-processing capabilities it provides. Here, we show that the BMP pathway processes multi-ligand inputs using a specific repertoire of computations, including ratiometric sensing, balance detection, and imbalance detection. These computations operate on the relative levels of different ligands and can arise directly from competitive receptor-ligand interactions. Furthermore, cells can select different computations to perform on the same ligand combination through expression of alternative sets of receptor variants. These results provide a direct signal-processing role for promiscuous receptor-ligand interactions and establish operational principles for quantitatively controlling cells with BMP ligands. Similar principles could apply to other promiscuous signaling pathways.

Programmable removal of bacterial strains by use of genome-targeting CRISPR-Cas systems.

UNLABELLED: CRISPR (clustered regularly interspaced short palindromic repeats)-Cas (CRISPR-associated) systems in bacteria and archaea employ CRISPR RNAs to specifically recognize the complementary DNA of foreign invaders, leading to sequence-specific cleavage or degradation of the target DNA. Recent work has shown that the accidental or intentional targeting of the bacterial genome is cytotoxic and can lead to cell death. Here, we have demonstrated that genome targeting with CRISPR-Cas systems can be employed for the sequence-specific and titratable removal of individual bacterial strains and species. Using the type I-E CRISPR-Cas system in Escherichia coli as a model, we found that this effect could be elicited using native or imported systems and was similarly potent regardless of the genomic location, strand, or transcriptional activity of the target sequence. Furthermore, the specificity of targeting with CRISPR RNAs could readily distinguish between even highly similar strains in pure or mixed cultures. Finally, varying the collection of delivered CRISPR RNAs could quantitatively control the relative number of individual strains within a mixed culture. Critically, the observed selectivity and programmability of bacterial removal would be virtually impossible with traditional antibiotics, bacteriophages, selectable markers, or tailored growth conditions. Once delivery challenges are addressed, we envision that this approach could offer a novel means to quantitatively control the composition of environmental and industrial microbial consortia and may open new avenues for the development of "smart" antibiotics that circumvent multidrug resistance and differentiate between pathogenic and beneficial microorganisms. IMPORTANCE: Controlling the composition of microbial populations is a critical aspect in medicine, biotechnology, and environmental cycles. While different antimicrobial strategies, such as antibiotics, antimicrobial peptides, and lytic bacteriophages, offer partial solutions, what remains elusive is a generalized and programmable strategy that can distinguish between even closely related microorganisms and that allows for fine control over the composition of a microbial population. This study demonstrates that RNA-directed immune systems in bacteria and archaea called CRISPR-Cas systems can provide such a strategy. These systems can be employed to selectively and quantitatively remove individual bacterial strains based purely on sequence information, creating opportunities in the treatment of multidrug-resistant infections, the control of industrial fermentations, and the study of microbial consortia.