Early risk assessment for COVID-19 patients from emergency department data using machine learning.

Since its emergence in late 2019, the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has caused a pandemic with more than 55 million reported cases and 1.3 million estimated deaths worldwide. While epidemiological and clinical characteristics of COVID-19 have been reported, risk factors underlying the transition from mild to severe disease among patients remain poorly understood. In this retrospective study, we analysed data of 879 confirmed SARS-CoV-2 positive patients admitted to a two-site NHS Trust hospital in London, England, between January 1st and May 26th, 2020, with a majority of cases occurring in March and April. We extracted anonymised demographic data, physiological clinical variables and laboratory results from electronic healthcare records (EHR) and applied multivariate logistic regression, random forest and extreme gradient boosted trees. To evaluate the potential for early risk assessment, we used data available during patients' initial presentation at the emergency department (ED) to predict deterioration to one of three clinical endpoints in the remainder of the hospital stay: admission to intensive care, need for invasive mechanical ventilation and in-hospital mortality. Based on the trained models, we extracted the most informative clinical features in determining these patient trajectories. Considering our inclusion criteria, we have identified 129 of 879 (15%) patients that required intensive care, 62 of 878 (7%) patients needing mechanical ventilation, and 193 of 619 (31%) cases of in-hospital mortality. Our models learned successfully from early clinical data and predicted clinical endpoints with high accuracy, the best model achieving area under the receiver operating characteristic (AUC-ROC) scores of 0.76 to 0.87 (F1 scores of 0.42-0.60). Younger patient age was associated with an increased risk of receiving intensive care and ventilation, but lower risk of mortality. Clinical indicators of a patient's oxygen supply and selected laboratory results, such as blood lactate and creatinine levels, were most predictive of COVID-19 patient trajectories. Among COVID-19 patients machine learning can aid in the early identification of those with a poor prognosis, using EHR data collected during a patient's first presentation at ED. Patient age and measures of oxygenation status during ED stay are primary indicators of poor patient outcomes.

A Single Light-Responsive Sizer Can Control Multiple-Fission Cycles in Chlamydomonas.

Most eukaryotic cells execute binary division after each mass doubling in order to maintain size homeostasis by coordinating cell growth and division. By contrast, the photosynthetic green alga Chlamydomonas can grow more than 8-fold during daytime and then, at night, undergo rapid cycles of DNA replication, mitosis, and cell division, producing up to 16 daughter cells. Here, we propose a mechanistic model for multiple-fission cycles and cell-size control in Chlamydomonas. The model comprises a light-sensitive and size-dependent biochemical toggle switch that acts as a sizer, guarding transitions into and exit from a phase of cell-division cycle oscillations. This simple "sizer-oscillator" arrangement reproduces the experimentally observed features of multiple-fission cycles and the response of Chlamydomonas cells to different light-dark regimes. Our model also makes specific predictions about the size dependence of the time of onset of cell division after cells are transferred from light to dark conditions, and we confirm these predictions by single-cell experiments. Collectively, our results provide a new perspective on the concept of a "commitment point" during the growth of Chlamydomonas cells and hint at intriguing similarities of cell-size control in different eukaryotic lineages.

Dilution and titration of cell-cycle regulators may control cell size in budding yeast.

The size of a cell sets the scale for all biochemical processes within it, thereby affecting cellular fitness and survival. Hence, cell size needs to be kept within certain limits and relatively constant over multiple generations. However, how cells measure their size and use this information to regulate growth and division remains controversial. Here, we present two mechanistic mathematical models of the budding yeast (S. cerevisiae) cell cycle to investigate competing hypotheses on size control: inhibitor dilution and titration of nuclear sites. Our results suggest that an inhibitor-dilution mechanism, in which cell growth dilutes the transcriptional inhibitor Whi5 against the constant activator Cln3, can facilitate size homeostasis. This is achieved by utilising a positive feedback loop to establish a fixed size threshold for the Start transition, which efficiently couples cell growth to cell cycle progression. Yet, we show that inhibitor dilution cannot reproduce the size of mutants that alter the cell's overall ploidy and WHI5 gene copy number. By contrast, size control through titration of Cln3 against a constant number of genomic binding sites for the transcription factor SBF recapitulates both size homeostasis and the size of these mutant strains. Moreover, this model produces an imperfect 'sizer' behaviour in G1 and a 'timer' in S/G2/M, which combine to yield an 'adder' over the whole cell cycle; an observation recently made in experiments. Hence, our model connects these phenomenological data with the molecular details of the cell cycle, providing a systems-level perspective of budding yeast size control.

A comprehensive model for the proliferation-quiescence decision in response to endogenous DNA damage in human cells.

Human cells that suffer mild DNA damage can enter a reversible state of growth arrest known as quiescence. This decision to temporarily exit the cell cycle is essential to prevent the propagation of mutations, and most cancer cells harbor defects in the underlying control system. Here we present a mechanistic mathematical model to study the proliferation-quiescence decision in nontransformed human cells. We show that two bistable switches, the restriction point (RP) and the G1/S transition, mediate this decision by integrating DNA damage and mitogen signals. In particular, our data suggest that the cyclin-dependent kinase inhibitor p21 (Cip1/Waf1), which is expressed in response to DNA damage, promotes quiescence by blocking positive feedback loops that facilitate G1 progression downstream of serum stimulation. Intriguingly, cells exploit bistability in the RP to convert graded p21 and mitogen signals into an all-or-nothing cell-cycle response. The same mechanism creates a window of opportunity where G1 cells that have passed the RP can revert to quiescence if exposed to DNA damage. We present experimental evidence that cells gradually lose this ability to revert to quiescence as they progress through G1 and that the onset of rapid p21 degradation at the G1/S transition prevents this response altogether, insulating S phase from mild, endogenous DNA damage. Thus, two bistable switches conspire in the early cell cycle to provide both sensitivity and robustness to external stimuli.

DNA damage during S-phase mediates the proliferation-quiescence decision in the subsequent G1 via p21 expression.

Following DNA damage caused by exogenous sources, such as ionizing radiation, the tumour suppressor p53 mediates cell cycle arrest via expression of the CDK inhibitor, p21. However, the role of p21 in maintaining genomic stability in the absence of exogenous DNA-damaging agents is unclear. Here, using live single-cell measurements of p21 protein in proliferating cultures, we show that naturally occurring DNA damage incurred over S-phase causes p53-dependent accumulation of p21 during mother G2- and daughter G1-phases. High p21 levels mediate G1 arrest via CDK inhibition, yet lower levels have no impact on G1 progression, and the ubiquitin ligases CRL4Cdt2 and SCFSkp2 couple to degrade p21 prior to the G1/S transition. Mathematical modelling reveals that a bistable switch, created by CRL4Cdt2, promotes irreversible S-phase entry by keeping p21 levels low, preventing premature S-phase exit upon DNA damage. Thus, we characterize how p21 regulates the proliferation-quiescence decision to maintain genomic stability.

A Dynamical Framework for the All-or-None G1/S Transition.

The transition from G1 into DNA replication (S phase) is an emergent behavior resulting from dynamic and complex interactions between cyclin-dependent kinases (Cdks), Cdk inhibitors (CKIs), and the anaphase-promoting complex/cyclosome (APC/C). Understanding the cellular decision to commit to S phase requires a quantitative description of these interactions. We apply quantitative imaging of single human cells to track the expression of G1/S regulators and use these data to parametrize a stochastic mathematical model of the G1/S transition. We show that a rapid, proteolytic, double-negative feedback loop between Cdk2:Cyclin and the Cdk inhibitor p27(Kip1) drives a switch-like entry into S phase. Furthermore, our model predicts that increasing Emi1 levels throughout S phase are critical in maintaining irreversibility of the G1/S transition, which we validate using Emi1 knockdown and live imaging of G1/S reporters. This work provides insight into the general design principles of the signaling networks governing the temporally abrupt transitions between cell-cycle phases.

Community control in cellular protein production: consequences for amino acid starvation.

Deprivation of essential nutrients can have stark consequences for many processes in a cell. We consider amino acid starvation, which can result in bottlenecks in mRNA translation when ribosomes stall due to lack of resources, i.e. tRNAs charged with the missing amino acid. Recent experiments also show less obvious effects such as increased charging of other (non-starved) tRNA species and selective charging of isoaccepting tRNAs. We present a mechanism which accounts for these observations and shows that production of some proteins can actually increase under starvation. One might assume that such responses could only be a result of sophisticated control pathways, but here we show that these effects can occur naturally due to changes in the supply and demand for different resources, and that control can be accomplished through selective use of rare codons. We develop a model for translation which includes the dynamics of the charging and use of aminoacylated tRNAs, explicitly taking into account the effect of specific codon sequences. This constitutes a new control mechanism in gene regulation which emerges at the community level, i.e. via resources used by all ribosomes.

Single-cell analysis and stochastic modelling unveil large cell-to-cell variability in influenza A virus infection.

Biochemical reactions are subject to stochastic fluctuations that can give rise to cell-to-cell variability. Yet, how this variability affects viral infections, which themselves involve noisy reactions, remains largely elusive. Here we present single-cell experiments and stochastic simulations that reveal a large heterogeneity between influenza A virus (IAV)-infected cells. In particular, experimental data show that progeny virus titres range from 1 to 970 plaque-forming units and intracellular viral RNA (vRNA) levels span three orders of magnitude. Moreover, the segmentation of IAV genomes seems to increase the susceptibility of their replication to noise, since the level of different genome segments can vary substantially within a cell. In addition, simulations suggest that the abortion of virus entry and random degradation of vRNAs can result in a large fraction of non-productive cells after single-hit infection. These results challenge current beliefs that cell population measurements and deterministic simulations are an accurate representation of viral infections.

Multiscale modeling of influenza A virus infection supports the development of direct-acting antivirals.

Influenza A viruses are respiratory pathogens that cause seasonal epidemics with up to 500,000 deaths each year. Yet there are currently only two classes of antivirals licensed for treatment and drug-resistant strains are on the rise. A major challenge for the discovery of new anti-influenza agents is the identification of drug targets that efficiently interfere with viral replication. To support this step, we developed a multiscale model of influenza A virus infection which comprises both the intracellular level where the virus synthesizes its proteins, replicates its genome, and assembles new virions and the extracellular level where it spreads to new host cells. This integrated modeling approach recapitulates a wide range of experimental data across both scales including the time course of all three viral RNA species inside an infected cell and the infection dynamics in a cell population. It also allowed us to systematically study how interfering with specific steps of the viral life cycle affects virus production. We find that inhibitors of viral transcription, replication, protein synthesis, nuclear export, and assembly/release are most effective in decreasing virus titers whereas targeting virus entry primarily delays infection. In addition, our results suggest that for some antivirals therapy success strongly depends on the lifespan of infected cells and, thus, on the dynamics of virus-induced apoptosis or the host's immune response. Hence, the proposed model provides a systems-level understanding of influenza A virus infection and therapy as well as an ideal platform to include further levels of complexity toward a comprehensive description of infectious diseases.

Modeling the intracellular dynamics of influenza virus replication to understand the control of viral RNA synthesis.

Influenza viruses transcribe and replicate their negative-sense RNA genome inside the nucleus of host cells via three viral RNA species. In the course of an infection, these RNAs show distinct dynamics, suggesting that differential regulation takes place. To investigate this regulation in a systematic way, we developed a mathematical model of influenza virus infection at the level of a single mammalian cell. It accounts for key steps of the viral life cycle, from virus entry to progeny virion release, while focusing in particular on the molecular mechanisms that control viral transcription and replication. We therefore explicitly consider the nuclear export of viral genome copies (vRNPs) and a recent hypothesis proposing that replicative intermediates (cRNA) are stabilized by the viral polymerase complex and the nucleoprotein (NP). Together, both mechanisms allow the model to capture a variety of published data sets at an unprecedented level of detail. Our findings provide theoretical support for an early regulation of replication by cRNA stabilization. However, they also suggest that the matrix protein 1 (M1) controls viral RNA levels in the late phase of infection as part of its role during the nuclear export of viral genome copies. Moreover, simulations show an accumulation of viral proteins and RNA toward the end of infection, indicating that transport processes or budding limits virion release. Thus, our mathematical model provides an ideal platform for a systematic and quantitative evaluation of influenza virus replication and its complex regulation.