Collective Space-Sensing Coordinates Pattern Scaling in Engineered Bacteria.

Scale invariance refers to the maintenance of a constant ratio of developing organ size to body size. Although common, its underlying mechanisms remain poorly understood. Here, we examined scaling in engineered Escherichia coli that can form self-organized core-ring patterns in colonies. We found that the ring width exhibits perfect scale invariance to the colony size. Our analysis revealed a collective space-sensing mechanism, which entails sequential actions of an integral feedback loop and an incoherent feedforward loop. The integral feedback is implemented by the accumulation of a diffusive chemical produced by a colony. This accumulation, combined with nutrient consumption, sets the timing for ring initiation. The incoherent feedforward is implemented by the opposing effects of the domain size on the rate and duration of ring maturation. This mechanism emphasizes a role of timing control in achieving robust pattern scaling and provides a new perspective in examining the phenomenon in natural systems.

Fully connected autoencoder and convolutional neural network with attention-based method for inferring disease-related lncRNAs.

Since abnormal expression of long noncoding RNAs (lncRNAs) is often closely related to various human diseases, identification of disease-associated lncRNAs is helpful for exploring the complex pathogenesis. Most of recent methods concentrate on exploiting multiple kinds of data related to lncRNAs and diseases for predicting candidate disease-related lncRNAs. These methods, however, failed to deeply integrate the topology information from the meta-paths that are composed of lncRNA, disease and microRNA (miRNA) nodes. We proposed a new method based on fully connected autoencoders and convolutional neural networks, called ACLDA, for inferring potential disease-related lncRNA candidates. A heterogeneous graph that consists of lncRNA, disease and miRNA nodes were firstly constructed to integrate similarities, associations and interactions among them. Fully connected autoencoder-based module was established to extract the low-dimensional features of lncRNA, disease and miRNA nodes in the heterogeneous graph. We designed the attention mechanisms at the node feature level and at the meta-path level to learn more informative features and meta-paths. A module based on convolutional neural networks was constructed to encode the local topologies of lncRNA and disease nodes from multiple meta-path perspectives. The comprehensive experimental results demonstrated ACLDA achieves superior performance than several state-of-the-art prediction methods. Case studies on breast, lung and colon cancers demonstrated that ACLDA is able to discover the potential disease-related lncRNAs.

Stochastic sensitivity analysis and kernel inference via distributional data.

Cellular processes are noisy due to the stochastic nature of biochemical reactions. As such, it is impossible to predict the exact quantity of a molecule or other attributes at the single-cell level. However, the distribution of a molecule over a population is often deterministic and is governed by the underlying regulatory networks relevant to the cellular functionality of interest. Recent studies have started to exploit this property to infer network states. To facilitate the analysis of distributional data in a general experimental setting, we introduce a computational framework to efficiently characterize the sensitivity of distributional output to changes in external stimuli. Further, we establish a probability-divergence-based kernel regression model to accurately infer signal level based on distribution measurements. Our methodology is applicable to any biological system subject to stochastic dynamics and can be used to elucidate how population-based information processing may contribute to organism-level functionality. It also lays the foundation for engineering synthetic biological systems that exploit population decoding to more robustly perform various biocomputation tasks, such as disease diagnostics and environmental-pollutant sensing.

Temporal control of self-organized pattern formation without morphogen gradients in bacteria.

Diverse mechanisms have been proposed to explain biological pattern formation. Regardless of their specific molecular interactions, the majority of these mechanisms require morphogen gradients as the spatial cue, which are either predefined or generated as a part of the patterning process. However, using Escherichia coli programmed by a synthetic gene circuit, we demonstrate here the generation of robust, self-organized ring patterns of gene expression in the absence of an apparent morphogen gradient. Instead of being a spatial cue, the morphogen serves as a timing cue to trigger the formation and maintenance of the ring patterns. The timing mechanism enables the system to sense the domain size of the environment and generate patterns that scale accordingly. Our work defines a novel mechanism of pattern formation that has implications for understanding natural developmental processes.

CLN-617 Retains IL2 and IL12 in Injected Tumors to Drive Robust and Systemic Immune-Mediated Antitumor Activity.

Despite clinical evidence of antitumor activity, the development of cytokine therapies has been hampered by a narrow therapeutic window and limited response rates. Two cytokines of high interest for clinical development are interleukin 2 (IL2) and interleukin 12 (IL12), which potently synergize to promote the activation and proliferation of T cells and NK cells. However, the only approved human IL2 therapy, Proleukin, is rarely used in the clinic due to systemic toxicities, and no IL12 product has been approved to date due to severe dose-limiting toxicities. Here, we describe CLN-617, a first-in-class therapeutic for intratumoral (IT) injection that co-delivers IL2 and IL12 on a single molecule in a safe and effective manner. CLN-617 is a single-chain fusion protein comprised of IL2, leukocyte-associated immunoglobulin-like receptor 2 (LAIR2), human serum albumin (HSA), and IL12. LAIR2 and HSA function to retain CLN-617 in the treated tumor by binding collagen and increasing molecular weight, respectively. We found that IT administration of a murine surrogate of CLN-617, mCLN-617, eradicated established treated and untreated tumors in syngeneic models, significantly improved response to anti-PD1 checkpoint therapy, and generated a robust abscopal response dependent on cellular immunity and antigen cross-presentation. CLN-617 is being evaluated in a clinical trial in patients with advanced solid tumors (NCT06035744).

Tension and robustness in multitasking cellular networks.

Cellular networks multitask by exhibiting distinct, context-dependent dynamics. However, network states (parameters) that generate a particular dynamic are often sub-optimal for others, defining a source of "tension" between them. Though multitasking is pervasive, it is not clear where tension arises, what consequences it has, and how it is resolved. We developed a generic computational framework to examine the source and consequences of tension between pairs of dynamics exhibited by the well-studied RB-E2F switch regulating cell cycle entry. We found that tension arose from task-dependent shifts in parameters associated with network modules. Although parameter sets common to distinct dynamics did exist, tension reduced both their accessibility and resilience to perturbation, indicating a trade-off between "one-size-fits-all" solutions and robustness. With high tension, robustness can be preserved by dynamic shifting of modules, enabling the network to toggle between tasks, and by increasing network complexity, in this case by gene duplication. We propose that tension is a general constraint on the architecture and operation of multitasking biological networks. To this end, our work provides a framework to quantify the extent of tension between any network dynamics and how it affects network robustness. Such analysis would suggest new ways to interfere with network elements to elucidate the design principles of cellular networks.

Comparative analysis of alignment algorithms for macular optical coherence tomography imaging.

BACKGROUND: Optical coherence tomography (OCT) is the most important and commonly utilized imaging modality in ophthalmology and is especially crucial for the diagnosis and management of macular diseases. Each OCT volume is typically only available as a series of cross-sectional images (B-scans) that are accessible through proprietary software programs which accompany the OCT machines. To maximize the potential of OCT imaging for machine learning purposes, each OCT image should be analyzed en bloc as a 3D volume, which requires aligning all the cross-sectional images within a particular volume. METHODS: A dataset of OCT B-scans obtained from 48 age-related macular degeneration (AMD) patients and 50 normal controls was used to evaluate five registration algorithms. After alignment of B-scans from each patient, an en face surface map was created to measure the registration quality, based on an automatically generated Laplace difference of the surface map-the smoother the surface map, the smaller the average Laplace difference. To demonstrate the usefulness of B-scan alignment, we trained a 3D convolutional neural network (CNN) to detect age-related macular degeneration (AMD) on OCT images and compared the performance of the model with and without B-scan alignment. RESULTS: The mean Laplace difference of the surface map before registration was 27 ± 4.2 pixels for the AMD group and 26.6 ± 4 pixels for the control group. After alignment, the smoothness of the surface map was improved, with a mean Laplace difference of 5.5 ± 2.7 pixels for Advanced Normalization Tools Symmetric image Normalization (ANTs-SyN) registration algorithm in the AMD group and a mean Laplace difference of 4.3 ± 1.4.2 pixels for ANTs in the control group. Our 3D CNN achieved superior performance in detecting AMD, when aligned OCT B-scans were used (AUC 0.95 aligned vs. 0.89 unaligned). CONCLUSIONS: We introduced a novel metric to quantify OCT B-scan alignment and compared the effectiveness of five alignment algorithms. We confirmed that alignment could be improved in a statistically significant manner with readily available alignment algorithms that are available to the public, and the ANTs algorithm provided the most robust performance overall. We further demonstrated that alignment of OCT B-scans will likely be useful for training 3D CNN models.

CLN-978, a novel half-life extended CD19/CD3/HSA-specific T cell-engaging antibody construct with potent activity against B-cell malignancies with low CD19 expression.

BACKGROUND: Despite significant progress in the development of T cell-engaging therapies for various B-cell malignancies, a high medical need remains for the refractory disease setting, often characterized by suboptimal target levels. METHODS: To address this issue, we have developed a 65-kDa multispecific antibody construct, CLN-978, with affinities tuned to optimize the killing of low-CD19 expressing tumor cells. CLN-978 bound to CD19 on B cells with picomolar affinity, and to CD3ε on T cells with nanomolar affinity. A serum albumin binding domain was incorporated to extend serum half-life. In this setting, we biophysically characterize and report the activities of CLN-978 in cell co-culture assays, multiple mouse models and non-human primates. RESULTS: Human T cells redirected by CLN-978 could eliminate target cells expressing less than 300 copies of CD19 on their surface. The half-life extension and high affinity for CD19 led to significant antitumor activity in murine lymphoma models at very low doses of CLN-978. In primates, we observed a long serum half-life, deep and sustained depletion of normal B cells, and remarkable tolerability, in particular, reduced cytokine release when CLN-978 was administered subcutaneously. CONCLUSIONS: CLN-978 warrants further exploration. An ongoing clinical phase 1 trial is investigating safety, pharmacokinetics, pharmacodynamics, and the initial therapeutic potential of subcutaneously administered CLN-978 in patients with non-Hodgkin's lymphoma.

Computation of steady-state probability distributions in stochastic models of cellular networks.

Cellular processes are "noisy". In each cell, concentrations of molecules are subject to random fluctuations due to the small numbers of these molecules and to environmental perturbations. While noise varies with time, it is often measured at steady state, for example by flow cytometry. When interrogating aspects of a cellular network by such steady-state measurements of network components, a key need is to develop efficient methods to simulate and compute these distributions. We describe innovations in stochastic modeling coupled with approaches to this computational challenge: first, an approach to modeling intrinsic noise via solution of the chemical master equation, and second, a convolution technique to account for contributions of extrinsic noise. We show how these techniques can be combined in a streamlined procedure for evaluation of different sources of variability in a biochemical network. Evaluation and illustrations are given in analysis of two well-characterized synthetic gene circuits, as well as a signaling network underlying the mammalian cell cycle entry.