Search

AlphaFold2 and related computational systems predict protein structure using deep learning and co-evolutionary relationships encoded in multiple sequence alignments (MSAs). Despite high prediction accuracy achieved by these systems, challenges remain in (1) prediction of orphan and rapidly evolving proteins for which an MSA cannot be generated; (2) rapid exploration of designed structures; and (3) understanding the rules governing spontaneous polypeptide folding in solution. Here we report development of an end-to-end differentiable recurrent geometric network (RGN) that uses a protein language model (AminoBERT) to learn latent structural information from unaligned proteins. A linked geometric module compactly represents C

Deep learning using neural networks relies on a class of machine-learnable models constructed using 'differentiable programs'. These programs can combine mathematical equations specific to a particular domain of natural science with general-purpose, machine-learnable components trained on experimental data. Such programs are having a growing impact on molecular and cellular biology. In this Perspective, we describe an emerging 'differentiable biology' in which phenomena ranging from the small and specific (for example, one experimental assay) to the broad and complex (for example, protein folding) can be modeled effectively and efficiently, often by exploiting knowledge about basic natural phenomena to overcome the limitations of sparse, incomplete and noisy data. By distilling differentiable biology into a small set of conceptual primitives and illustrative vignettes, we show how it can help to address long-standing challenges in integrating multimodal data from diverse experiments across biological scales. This promises to benefit fields as diverse as biophysics and functional genomics.

In mammalian cells, much of signal transduction is mediated by weak protein-protein interactions between globular peptide-binding domains (PBDs) and unstructured peptidic motifs in partner proteins. The number and diversity of these PBDs (over 1,800 are known), their low binding affinities and the sensitivity of binding properties to minor sequence variation represent a substantial challenge to experimental and computational analysis of PBD specificity and the networks PBDs create. Here, we introduce a bespoke machine-learning approach, hierarchical statistical mechanical modeling (HSM), capable of accurately predicting the affinities of PBD-peptide interactions across multiple protein families. By synthesizing biophysical priors within a modern machine-learning framework, HSM outperforms existing computational methods and high-throughput experimental assays. HSM models are interpretable in familiar biophysical terms at three spatial scales: the energetics of protein-peptide binding, the multidentate organization of protein-protein interactions and the global architecture of signaling networks.

Individual participant data (IPD) from oncology clinical trials is invaluable for identifying factors that influence trial success and failure, improving trial design and interpretation, and comparing pre-clinical studies to clinical outcomes. However, the IPD used to generate published survival curves are not generally publicly available. We impute survival IPD from ~500 arms of Phase 3 oncology trials (representing ~220,000 events) and find that they are well fit by a two-parameter Weibull distribution. Use of Weibull functions with overall survival significantly increases the precision of small arms typical of early phase trials: analysis of a 50-patient trial arm using parametric forms is as precise as traditional, non-parametric analysis of a 90-patient arm. We also show that frequent deviations from the Cox proportional hazards assumption, particularly in trials of immune checkpoint inhibitors, arise from time-dependent therapeutic effects. Trial duration therefore has an underappreciated impact on the likelihood of success.

The inflamed rheumatic joint is a highly heterogeneous and complex tissue with dynamic recruitment and expansion of multiple cell types that interact in multifaceted ways within a localized area. Rheumatoid arthritis synovium has primarily been studied either by immunostaining or by molecular profiling after tissue homogenization. Here, we use Spatial Transcriptomics, where tissue-resident RNA is spatially labeled in situ with barcodes in a transcriptome-wide fashion, to study local tissue interactions at the site of chronic synovial inflammation. We report comprehensive spatial RNA-Seq data coupled to cell type-specific localization patterns at and around organized structures of infiltrating leukocyte cells in the synovium. Combining morphological features and high-throughput spatially resolved transcriptomics may be able to provide higher statistical power and more insights into monitoring disease severity and treatment-specific responses in seropositive and seronegative rheumatoid arthritis.

Background The COVID-19 pandemic has severely disrupted supply chains for many types of Personal Protective Equipment (PPE), particularly surgical N95 filtering facepiece respirators (FFRs; “masks”). As a consequence, an Emergency Use Authorization (EUA) from the FDA has allowed use of industrial N95 respirators and importation of N95-type masks manufactured to international standards; these include KN95 masks from China and FFP2 masks from the European Union. Methods We conducted a survey of masks in the inventory of major academic medical centers in Boston, MA to determine provenance and manufacturer or supplier. We then assembled a testing apparatus at a university laboratory and performed a modified test of filtration performance using KCl and ambient particulate matter on masks from hospital inventories; an accompanying website shows how to build and use the testing apparatus. Results Over 100 different makes and models of traditional and nontraditional filtering facepiece respirators (N95-type masks) were in the inventory of surveyed U.S. teaching hospitals as opposed to 2–5 models under normal circumstances. A substantial number of unfamiliar masks are from unknown manufacturers. Many are not correctly labelled and do not perform to accepted standards and a subset are obviously dangerous; many of these masks are likely to be counterfeit. Due to the absence of publicly available information on mask suppliers and inconsistent labeling of KN95 masks, it is difficult to distinguish between legitimate and counterfeit products. Conclusions Many FFRs available for procurement during the COVID-19 pandemic do not provide levels of fit and filtration similar to those of N95 masks and are not acceptable for use in healthcare settings. Based on these results, and in consultation with occupational health officers, we make six recommendations to assist end users in acquiring legitimate products. Institutions should always assess masks from non-traditional supply chains by checking their markings and manufacturer information against data provided by NIOSH and the latest FDA EUA Appendix A. In the absence of verifiable information on the legitimacy of mask source, institutions should consider measuring mask fit and filtration directly. We also make suggestions for regulatory agencies regarding labeling and public disclosure aimed at increasing pandemic resilience.

Background In response to supply shortages caused by the COVID-19 pandemic, N95 filtering facepiece respirators (FFRs or “masks”), which are typically single-use devices in healthcare settings, are routinely being used for prolonged periods and in some cases decontaminated under “reuse” and “extended use” policies. However, the reusability of N95 masks is limited by degradation of fit. Possible substitutes, such as KN95 masks meeting Chinese standards, frequently fail fit testing even when new. The purpose of this study was to develop an inexpensive frame for damaged and poorly fitting masks using readily available materials and 3D printing. Results An iterative design process yielded a mask frame consisting of two 3D printed side pieces, malleable wire links that users press against their face, and cut lengths of elastic material that go around the head to hold the frame and mask in place. Volunteers (n = 45; average BMI = 25.4), underwent qualitative fit testing with and without mask frames wearing one or more of four different brands of FFRs conforming to US N95 or Chinese KN95 standards. Masks passed qualitative fit testing in the absence of a frame at rates varying from 48 to 94 % (depending on mask model). For individuals who underwent testing using respirators with broken or defective straps, 80–100 % (average 85 %) passed fit testing with mask frames. Among individuals who failed fit testing with a KN95, ~ 50 % passed testing by using a frame. Conclusions Our study suggests that mask frames can prolong the lifespan of N95 and KN95 masks by serving as a substitute for broken or defective bands without adversely affecting fit. Use of frames made it possible for ~ 73 % of the test population to achieve a good fit based on qualitative and quantitative testing criteria, approaching the 85–90 % success rate observed for intact N95 masks. Frames therefore represent a simple and inexpensive way of expanding access to PPE and extending their useful life. For clinicians and institutions interested in mask frames, designs and specifications are provided without restriction for use or modification. To ensure adequate performance in clinical settings, fit testing with user-specific masks and PanFab frames is required.

Multiplexed tissue imaging enables precise, spatially resolved enumeration and characterization of cell types and states in human resection specimens. A growing number of methods applicable to formalin-fixed, paraffin-embedded (FFPE) tissue sections have been described, the majority of which rely on antibodies for antigen detection and mapping. This protocol provides step-by-step procedures for confirming the selectivity and specificity of antibodies used in fluorescence-based tissue imaging and for the construction and validation of antibody panels. Although the protocol is implemented using tissue-based cyclic immunofluorescence (t-CyCIF) as an imaging platform, these antibody-testing methods are broadly applicable. We demonstrate assembly of a 16-antibody panel for enumerating and localizing T cells and B cells, macrophages, and cells expressing immune checkpoint regulators. The protocol is accessible to individuals with experience in microscopy and immunofluorescence; some experience in computation is required for data analysis. A typical 30-antibody dataset for 20 FFPE slides can be generated within 2 weeks.

The COVID-19 pandemic has led to widespread shortages of personal protective equipment (PPE) for healthcare workers, including of N95 masks (filtering facepiece respirators; FFRs). These masks are intended for single use but their sterilization and subsequent reuse has the potential to substantially mitigate shortages. Here we investigate PPE sterilization using ionized hydrogen peroxide (iHP), generated by SteraMist equipment (TOMI; Frederick, MD), in a sealed environment chamber. The efficacy of sterilization by iHP was assessed using bacterial spores in biological indicator assemblies. After one or more iHP treatments, five models of N95 masks from three manufacturers were assessed for retention of function based on their ability to form an airtight seal (measured using a quantitative fit test) and filter aerosolized particles. Filtration testing was performed at a university lab and at a National Institute for Occupational Safety and Health (NIOSH) pre-certification laboratory. The data demonstrate that N95 masks sterilized using SteraMist iHP technology retain filtration efficiency up to ten cycles, the maximum number tested to date. A typical iHP environment chamber with a volume of ~ 80 m3 can treat ~ 7000 masks and other items (e.g. other PPE, iPADs), making this an effective approach for a busy medical center.

Activation of synovial fibroblasts (SFs) contributes to rheumatoid arthritis (RA) by damaging synovial membranes and generating inflammatory cytokines that recruit immune cells to the joint. In this paper we profile cytokine secretion by primary human SFs from healthy tissues and from donors with RA and show that SF activation by TNF, IL-1α, and polyinosinic-polycytidylic acid (Poly(I:C)) cause secretion of multiple cytokines found at high levels in RA synovial fluids. We used interaction multiple linear regression to quantify therapeutic and countertherapeutic drug effects across activators and donors and found that the ability of drugs to block SF activation was strongly dependent on the identity of the activating cytokine. (5z)-7-oxozeaenol (5ZO), a preclinical drug that targets transforming growth factor-β-activated kinase 1 (TAK1), was more effective at blocking SF activation across all contexts than the approved drug tofacitinib, which supports the development of molecules similar to 5ZO for use as RA therapeutics., National Institutes of Health (U.S.). Ruth L. Kirschstein National Research Service Award (5F32AR062931)

Traditional means for scoring the effects of anti-cancer drugs on the growth and survival of cell lines is based on relative cell number in drug-treated and control samples and is seriously confounded by unequal division rates arising from natural biological variation and differences in culture conditions. This problem can be overcome by computing drug sensitivity on a per-division basis. The normalized growth rate inhibition (GR) approach yields per-division metrics for drug potency (GR50) and efficacy (GRmax) that are analogous to the more familiar IC50 and Emax values. In this work, we report GR-based, proliferation-corrected, drug sensitivity metrics for ~4,700 pairs of breast cancer cell lines and perturbagens. Such data are broadly useful in understanding the molecular basis of therapeutic response and resistance. Here, we use them to investigate the relationship between different measures of drug sensitivity and conclude that drug potency and efficacy exhibit high variation that is only weakly correlated. To facilitate further use of these data, computed GR curves and metrics can be browsed interactively at http://www.GRbrowser.org/.

Signal transduction by receptor tyrosine kinases (RTKs) involves complex ligand- and time-dependent changes in conformation and modification state. High resolution structures are available for individual receptors dimers, but less is known about receptor clusters that form in plasma membranes composed of many different RTKs with the potential to interact. We report the use of multiplexed super-resolution imaging (Exchange-PAINT) followed by mean-shift clustering and random forest analysis to measure the precise distributions of five receptor tyrosine kinases (RTKs) from the ErbB, IGF-1R and Met families in breast cancer cells. We find that these receptors are intermixed nonhomogenously on the plasma membrane. Stimulation by EGF does not appear to induce a change in the density of EGFR in local clusters but instead results in formation of EGFR-Met and EGFR-ErbB3 associations; non-canonical EGFR-Met interactions are implicated in resistance to anti-cancer drugs but have not been previously detected in drug-naïve cells.

Functional interpretation of genomic variation is critical to understanding human disease, but it remains difficult to predict the effects of specific mutations on protein interaction networks and the phenotypes they regulate. We describe an analytical framework based on multiscale statistical mechanics that integrates genomic and biophysical data to model the human SH2-phosphoprotein network in normal and cancer cells. We apply our approach to data in The Cancer Genome Atlas (TCGA) and test model predictions experimentally. We find that mutations mapping to phosphoproteins often create new interactions but that mutations altering SH2 domains result almost exclusively in loss of interactions. Some of these mutations eliminate all interactions, but many cause more selective loss, thereby rewiring specific edges in highly connected subnetworks. Moreover, idiosyncratic mutations appear to be as functionally consequential as recurrent mutations. By synthesizing genomic, structural and biochemical data, our framework represents a new approach to the interpretation of genetic variation.

Measuring the states of cell signaling pathways in tumor samples promises to advance the understanding of oncogenesis and identify response biomarkers. Here, we describe the use of Reverse Phase Protein Arrays (RPPAs or RPLAs) to profile signaling proteins in 56 breast cancers and matched normal tissue. In RPPAs, hundreds to thousands of lysates are arrayed in dense regular grids and each grid is probed with a different antibody (100 in the current work, of which 71 yielded strong signals with breast tissue). Although RPPA technology is quite widely used, measuring changes in phosphorylation reflective of protein activation remains challenging. Using repeat deposition and well-validated antibodies, we show that diverse patterns of phosphorylation can be monitored in tumor samples and changes mapped onto signaling networks in a coherent fashion. The patterns are consistent with biomarker-based classification of breast cancers and known mechanisms of oncogenesis. We explore in detail one tumor-associated pattern that involves changes in the abundance of the Axl receptor tyrosine kinase (RTK) and phosphorylation of the cMet RTK. Both cMet and Axl have been implicated in breast cancer, or in resistance to anticancer drugs, but the two RTKs are not known to be linked functionally. Protein depletion and overexpression studies in a 'triple-negative' breast cell line reveal cross talk between Axl and cMet involving Axl-mediated modification of cMet, a requirement for cMet in efficient and timely signal transduction by the Axl ligand Gas6 and the potential for the two receptors to interact physically. These findings have potential therapeutic implications, as they imply that bi-specific receptor inhibitors (for example, ATP-competitive small-kinase inhibitors such as GSK1363089, BMS-777607 or MP470) may be more efficacious than the mono-specific therapeutic antibodies currently in development (for example, Onartuzumab).

Variability in patient responses to even the most potent and targeted therapeutics is now the primary challenge facing drug discovery and patient care, particularly in oncology and immune therapy. Variability with respect to mechanisms of induced resistance is observed both in drug-naive patients and among those who are initially responsive. Genomics has developed powerful tools for systematic interrogation of disease genotype and transcriptional states (particularly in cancer) and for correlation of these measures with parameters of disease such as histological diagnosis and outcome. In contrast, mechanistic preclinical studies remain relatively narrowly focused, leading to many apparent contradictions and poor understanding of the determinants of response. We describe the emergence of a systems pharmacology approach that is mechanistic, quantitative, probabilistic, and postgenomic and promises to do for mechanistic pharmacology what genomics is doing for correlative studies. We focus on studies in cell lines (which currently dominate mechanism-oriented analysis), but our arguments are equally valid for real tumors studied in short-term culture as xenografts and, perhaps some time in the future, in humans.

Noisy gene expression or unequal partition of molecules during cell division are increasingly recognized as key sources of non-genetic cell-to-cell heterogeneity but the consequences for disease progression and drug efficiency are little understood. Through single-cell imaging, Spencer et al. now show that pre-existing cell-to-cell differences in the levels of signalling proteins determine whether the addition of an external death signal will kill a cell by apoptosis or not — and how quickly it happens. The mechanism may explain the phenomenon of 'fractional killing', in which repeated rounds of chemotherapy kill some but not all cells in a tumour. From an evolutionary perspective, such systems-level phenotypic variation — not based on genetic or epigenetic modifications — offers wider adaptive potential to populations of living organisms. Noise in gene expression gives rise to cell-to-cell variability in protein concentrations and is increasingly recognized as a key source of non-genetic differences between cells. Through single cell imaging, it has now been possible to demonstrate that pre-existing differences in the levels of signalling proteins determine whether the addition of an external death signal will kill a cell or not—and how fast. This has implications for understanding 'fractional killing' of tumour cells after chemotherapy, in which some but not all tumour cells die. In microorganisms, noise in gene expression gives rise to cell-to-cell variability in protein concentrations1,2,3,4,5,6,7. In mammalian cells, protein levels also vary8,9,10 and individual cells differ widely in their responsiveness to uniform physiological stimuli11,12,13,14,15. In the case of apoptosis mediated by TRAIL (tumour necrosis factor (TNF)-related apoptosis-inducing ligand) it is common for some cells in a clonal population to die while others survive—a striking divergence in cell fate. Among cells that die, the time between TRAIL exposure and caspase activation is highly variable. Here we image sister cells expressing reporters of caspase activation and mitochondrial outer membrane permeabilization after exposure to TRAIL. We show that naturally occurring differences in the levels or states of proteins regulating receptor-mediated apoptosis are the primary causes of cell-to-cell variability in the timing and probability of death in human cell lines. Protein state is transmitted from mother to daughter, giving rise to transient heritability in fate, but protein synthesis promotes rapid divergence so that sister cells soon become no more similar to each other than pairs of cells chosen at random. Our results have implications for understanding ‘fractional killing’ of tumour cells after exposure to chemotherapy, and for variability in mammalian signal transduction in general.

Defects in chromosome–microtubule attachment trigger spindle-checkpoint activation and delay mitotic progression1,2. How microtubule attachment is sensed and integrated into the steps of checkpoint-signal amplification is poorly understood. In a functional genomic screen targeting human kinases and phosphatases, we identified a microtubule affinity-regulating kinase kinase, TAO1 (also known as MARKK) as an important regulator of mitotic progression, required for both chromosome congression and checkpoint-induced anaphase delay. TAO1 interacts with the checkpoint kinase BubR1 and promotes enrichment of the checkpoint protein Mad2 at sites of defective attachment, providing evidence for a regulatory step that precedes the proposed Mad2–Mad1 dependent checkpoint-signal amplification step3. We propose that the dual functions of TAO1 in regulating microtubule dynamics and checkpoint signalling may help to coordinate the establishment and monitoring of correct congression of chromosomes, thereby protecting genomic stability in human cells.

Systems biology, particularly of mammalian cells, is data starved. However, technologies are now in place to obtain rich data, in a form suitable for model construction and validation, that describes the activities, states and locations of cell-signalling molecules. The key is to use several measurement technologies simultaneously and, recognizing each of their limits, to assemble a self-consistent compendium of systematic data.

Physicochemical modelling of signal transduction links fundamental chemical and physical principles, prior knowledge about regulatory pathways, and experimental data of various types to create powerful tools for formalizing and extending traditional molecular and cellular biology.

Combinatorial control of biological processes, in which redundancy and multifunctionality are the norm, fundamentally limits the therapeutic index that can be achieved by even the most potent and highly selective drugs. Thus, it will almost certainly be necessary to use new 'targeted' pharmaceuticals in combinations. Multicomponent drugs are standard in cytotoxic chemotherapy, but their development has required arduous empirical testing. However, experimentally validated numerical models should greatly aid in the formulation of new combination therapies, particularly those tailored to the needs of specific patients. This perspective focuses on opportunities and challenges inherent in the application of mathematical modeling and systems approaches to pharmacology, specifically with respect to the idea of achieving combinatorial selectivity through use of multicomponent drugs.

Cells must pass the correct number of chromosomes to their progeny through the complex ballet of cell division. An unusual conformation-sensitive switch seems to maintain accurate chromosome segregation.

Pharmacokinetic analysis at the organ level provides insight into how drugs distribute throughout the body, but cannot explain how drugs work at the cellular level. Here we demonstrate in vivo single-cell pharmacokinetic imaging of PARP-1 inhibitors and model drug behaviour under varying conditions. We visualize intracellular kinetics of the PARP-1 inhibitor distribution in real time, showing that PARP-1 inhibitors reach their cellular target compartment, the nucleus, within minutes in vivo both in cancer and normal cells in various cancer models. We also use these data to validate predictive finite element modelling. Our theoretical and experimental data indicate that tumour cells are exposed to sufficiently high PARP-1 inhibitor concentrations in vivo and suggest that drug inefficiency is likely related to proteomic heterogeneity or insensitivity of cancer cells to DNA-repair inhibition. This suggests that single-cell pharmacokinetic imaging and derived modelling improve our understanding of drug action at single-cell resolution in vivo.

IN somatic cells, entry into mitosis depends on the completion of DNA synthesis. This dependency is established by S-phase feed-back controls that arrest cell division when damaged or unreplicated DNA is present1. In the fission yeast Schizosaccharomyces pombe, mutations that interfere with the phosphorylation of tyrosine 15 (Y15) of p34cdc2, the protein kinase subunit of maturation promoting factor, accelerate the entry into mitosis and abolish the ability of unreplicated DNA to arrest cells in G2 (ref. 2). Because the tyrosine phosphorylation of p34cdc2 is conserved in S. pombe3, Xenopus4, chicken5 and human6cells, the regulation of p34cdc2-Y15 phosphorylation could be a universal mechanism mediating the S-phase feedback control and regulating the initiation of mitosis7,8. We have investigated these phenomena in the budding yeast Saccharomyces cerevisiae. We report here that the CDC28 gene product (the S.cerevisiae homologue of cdc2) is phosphorylated on the equivalent tyrosine (Y19) during S phase but that mutations that prevent tyrosine phosphorylation do not lead to premature mitosis and do not abolish feedback controls. We have therefore demonstrated a mechanism that does not involve tyrosine phosphorylation of p34 by which cells arrest their division in response to the presence of unreplicated or damaged DNA. We speculate that this mechanism may not involve the inactivation of p34 catalytic activity.

Blood comprises the largest version of the human proteome1. Changes of plasma protein profiles can reflect physiological or pathological conditions associated with many human diseases, making blood the most important fluid for clinical diagnostics2-4. Nevertheless, only a handful of plasma proteins are utilized in routine clinical tests. This is due to a host of reasons, including the intrinsic complexity of the plasma proteome1, the heterogeneity of human diseases and the fast kinetics associated with protein degradation in sampled blood5. Simple technologies that can sensitively sample large numbers of proteins over broad concentration ranges, from small amounts of blood, and within minutes of sample collection, would assist in solving these problems. Herein, we report on an integrated microfluidic system, called the Integrated Blood Barcode Chip (IBBC). It enables on-chip blood separation and the rapid measurement of a panel of plasma proteins from small quantities of blood samples including a fingerprick of whole blood. This platform holds potential for inexpensive, non-invasive, and informative clinical diagnoses, particularly, for point-of-care.

The Open Microscopy Environment (OME) defines a data model and a software implementation to serve as an informatics framework for imaging in biological microscopy experiments, including representation of acquisition parameters, annotations and image analysis results. OME is designed to support high-content cell-based screening as well as traditional image analysis applications. The OME Data Model, expressed in Extensible Markup Language (XML) and realized in a traditional database, is both extensible and self-describing, allowing it to meet emerging imaging and analysis needs.

When cells are exposed to elevated temperatures, transcription of a small set of genes, the heat-shock genes, is activated. This response is mediated by a short DNA sequence, the heat-shock element (HSE), which is thought to be the binding site for a specific transcription factor. Studies with Drosophila show that this protein binds to HSEs only in heat-shocked cells, suggesting that changes in factor binding are responsible for gene activation. We have investigated the properties of HSE-binding proteins from yeast and HeLa cells. In HeLa cells, binding activity is present only after heat shock. In contrast, control and heat-shocked yeast cells yield the same amount of HSE-binding activity; however, the mobility of protein-HSE complexes on polyacrylamide gels is altered following heat shock. This mobility difference can be significantly reduced by treatment of crude extracts with phosphatase. We propose that the yeast heat-shock factor binds constitutively to DNA but only activates transcription after heat-induced phosphorylation.