Adversarial training improves model interpretability in single-cell RNA-seq analysis.

Motivation: Predictive computational models must be accurate, robust, and interpretable to be considered reliable in important areas such as biology and medicine. A sufficiently robust model should not have its output affected significantly by a slight change in the input. Also, these models should be able to explain how a decision is made to support user trust in the results. Efforts have been made to improve the robustness and interpretability of predictive computational models independently; however, the interaction of robustness and interpretability is poorly understood. Results: As an example task, we explore the computational prediction of cell type based on single-cell RNA-seq data and show that it can be made more robust by adversarially training a deep learning model. Surprisingly, we find this also leads to improved model interpretability, as measured by identifying genes important for classification using a range of standard interpretability methods. Our results suggest that adversarial training may be generally useful to improve deep learning robustness and interpretability and that it should be evaluated on a range of tasks. Availability and implementation: Our Python implementation of all analysis in this publication can be found at: https://github.com/MehrshadSD/robustness-interpretability. The analysis was conducted using numPy 0.2.5, pandas 2.0.3, scanpy 1.9.3, tensorflow 2.10.0, matplotlib 3.7.1, seaborn 0.12.2, sklearn 1.1.1, shap 0.42.0, lime 0.2.0.1, matplotlib_venn 0.11.9.

Influence of administration time and sex on natriuretic, diuretic, and kaliuretic effects of diuretics.

Sex differences in renal function and blood pressure have been widely described across many species. Blood pressure dips during sleep and peaks in the early morning. Similarly, glomerular filtration rate, filtered electrolyte loads, urine volume, and urinary excretion all exhibit notable diurnal rhythms, which reflect, in part, the regulation of renal transporter proteins by circadian clock genes. That regulation is sexually dimorphic; as such, sex and time of day are not two independent regulators of kidney function and blood pressure. The objective of the present study was to assess the effect of sex and administration time on the natriuretic and diuretic effects of loop, thiazide, and K+-sparing diuretics, which are common treatments for hypertension. Loop diuretics inhibit Na+-K+-2Cl- cotransporters on the apical membrane of the thick ascending limb, thiazide diuretics inhibit Na+-Cl- cotransporters on the distal convoluted tubule, and K+-sparing diuretics inhibit epithelial Na+ channels on the connecting tubule and collecting duct. We simulated Na+ transporter inhibition using sex- and time-of-day-specific computational models of mouse kidney function. The simulation results highlighted significant sex and time-of-day differences in the drug response. Loop diuretics induced larger natriuretic and diuretic effects during the active phase. The natriuretic and diuretic effects of thiazide diuretics exhibited sex and time-of-day differences, whereas these effects of K+-sparing diuretics exhibited a significant time-of-day difference in females only. The kaliuretic effect depended on the type of diuretics and time of administration. The present computational models can be a useful tool in chronotherapy, to tailor drug administration time to match the body's diurnal rhythms to optimize the drug effect.NEW & NOTEWORTHY Sex influences cardiovascular disease, and the timing of onset of acute cardiovascular events exhibits circadian rhythms. Kidney function also exhibits sex differences and circadian rhythms. How do the natriuretic and diuretic effects of diuretics, a common treatment for hypertension that targets the kidneys, differ between the sexes? And how do these effects vary during the day? To answer these questions, we conducted computer simulations to assess the effects of loop, thiazide, and K+-sparing diuretics.

Understanding the dynamics of SARS-CoV-2 variants of concern in Ontario, Canada: a modeling study.

A year after the initial wild-type Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) strains began their devastation of the world, they were supplanted by new variants of concern (VOC). In Ontario, Canada, the wild type was overtaken first by the Alpha/B1.1.17 variant, and then by the Delta/B.1.617 variant. The principal objective of the present study is to develop and apply a much expanded Susceptible-Infection-Recovered-type model to better understand the spread of multiple VOC, and assess the effectiveness of vaccination and non-pharmaceutical interventions (NPI). The model represents competition among VOC, and reveals their mutual inhibitory effects. By separately tracking asymptomatic and symptomatic infections, model simulations identify a significant role of vaccine breakthrough in the spread of Delta. Furthermore, the severity of a Delta outbreak depends not only on the NPI and vaccination rate but also on the vaccine types. Alarmingly, despite Ontario's existing NPI and relatively successful vaccine rollout, a future, more dangerous VOC could potentially infect a significant fraction of the province's population and overwhelm the health care system. To stop that VOC, the province may need the simultaneous and rapid deployment of a third booster vaccine and stringent NPI.

The mixed blessing of AMPK signaling in Cancer treatments.

BACKGROUND: Nutrient acquisition and metabolism pathways are altered in cancer cells to meet bioenergetic and biosynthetic demands. A major regulator of cellular metabolism and energy homeostasis, in normal and cancer cells, is AMP-activated protein kinase (AMPK). AMPK influences cell growth via its modulation of the mechanistic target of Rapamycin (mTOR) pathway, specifically, by inhibiting mTOR complex mTORC1, which facilitates cell proliferation, and by activating mTORC2 and cell survival. Given its conflicting roles, the effects of AMPK activation in cancer can be counter intuitive. Prior to the establishment of cancer, AMPK acts as a tumor suppressor. However, following the onset of cancer, AMPK has been shown to either suppress or promote cancer, depending on cell type or state. METHODS: To unravel the controversial roles of AMPK in cancer, we developed a computational model to simulate the effects of pharmacological maneuvers that target key metabolic signalling nodes, with a specific focus on AMPK, mTORC, and their modulators. Specifically, we constructed an ordinary differential equation-based mechanistic model of AMPK-mTORC signaling, and parametrized the model based on existing experimental data. RESULTS: Model simulations were conducted to yield the following predictions: (i) increasing AMPK activity has opposite effects on mTORC depending on the nutrient availability; (ii) indirect inhibition of AMPK activity through inhibition of sirtuin 1 (SIRT1) only has an effect on mTORC activity under conditions of low nutrient availability; (iii) the balance between cell proliferation and survival exhibits an intricate dependence on DEP domain-containing mTOR-interacting protein (DEPTOR) abundance and AMPK activity; (iv) simultaneous direct inhibition of mTORC2 and activation of AMPK is a potential strategy for suppressing both cell survival and proliferation. CONCLUSIONS: Taken together, model simulations clarify the competing effects and the roles of key metabolic signalling pathways in tumorigenesis, which may yield insights on innovative therapeutic strategies.

Modeling within-Host SARS-CoV-2 Infection Dynamics and Potential Treatments.

The goal of this study was to develop a mathematical model to simulate the actions of drugs that target SARS-CoV-2 virus infection. To accomplish that goal, we have developed a mathematical model that describes the control of a SARS-CoV-2 infection by the innate and adaptive immune components. Invasion of the virus triggers the innate immunity, whereby interferon renders some of the target cells resistant to infection, and infected cells are removed by effector cells. The adaptive immune response is represented by plasma cells and virus-specific antibodies. The model is parameterized and then validated against viral load measurements collected in COVID-19 patients. We apply the model to simulate three potential anti-SARS-CoV-2 therapies: (1) Remdesivir, a repurposed drug that has been shown to inhibit the transcription of SARS-CoV-2, (2) an alternative (hypothetical) therapy that inhibits the virus' entry into host cells, and (3) convalescent plasma transfusion therapy. Simulation results point to the importance of early intervention, i.e., for any of the three therapies to be effective, it must be administered sufficiently early, not more than a day or two after the onset of symptoms. The model can serve as a key component in integrative platforms for rapid in silico testing of potential COVID-19 therapies and vaccines.

Interactions among mTORC, AMPK and SIRT: a computational model for cell energy balance and metabolism.

BACKGROUND: Cells adapt their metabolism and activities in response to signals from their surroundings, and this ability is essential for their survival in the face of perturbations. In tissues a deficit of these mechanisms is commonly associated with cellular aging and diseases, such as cardiovascular disease, cancer, immune system decline, and neurological pathologies. Several proteins have been identified as being able to respond directly to energy, nutrient, and growth factor levels and stress stimuli in order to mediate adaptations in the cell. In particular, mTOR, AMPK, and sirtuins are known to play an essential role in the management of metabolic stress and energy balance in mammals. METHODS: To understand the complex interactions of these signalling pathways and environmental signals, and how those interactions may impact lifespan and health-span, we have developed a computational model of metabolic signalling pathways. Specifically, the model includes (i) the insulin/IGF-1 pathway, which couples energy and nutrient abundance to the execution of cell growth and division, (ii) mTORC1 and the amino acid sensors such as sestrin, (iii) the Preiss-Handler and salvage pathways, which regulate the metabolism of NAD+ and the NAD+ -consuming factor SIRT1, (iv) the energy sensor AMPK, and (v) transcription factors FOXO and PGC-1α. RESULTS: The model simulates the interactions among key regulators such as AKT, mTORC1, AMPK, NAD+ , and SIRT, and predicts their dynamics. Key findings include the clinically important role of PRAS40 and diet in mTORC1 inhibition, and a potential link between SIRT1-activating compounds and premature autophagy. Moreover, the model captures the exquisite interactions of leucine, sestrin2, and arginine, and the resulting signal to the mTORC1 pathway. These results can be leveraged in the development of novel treatment of cancers and other diseases. CONCLUSIONS: This study presents a state-of-the-art computational model for investigating the interactions among signaling pathways and environmental stimuli in growth, ageing, metabolism, and diseases. The model can be used as an essential component to simulate gene manipulation, therapies (e.g., rapamycin and wortmannin), calorie restrictions, and chronic stress, and assess their functional implications on longevity and ageing-related diseases. Video Abstract.

Aging affects circadian clock and metabolism and modulates timing of medication.

Aging is associated with impairments in the circadian rhythms, and with energy deregulation that affects multiple metabolic pathways. The goal of this study is to unravel the complex interactions among aging, metabolism, and the circadian clock. We seek to identify key factors that inform the liver circadian clock of cellular energy status and to reveal the mechanisms by which variations in food intake may disrupt the clock. To address these questions, we develop a comprehensive mathematical model that represents the circadian pathway in the mouse liver, together with the insulin/IGF-1 pathway, mTORC1, AMPK, NAD+, and the NAD+ -consuming factor SIRT1. The model is age-specific and can simulate the liver of a young mouse or an aged mouse. Simulation results suggest that the reduced NAD+ and SIRT1 bioavailability may explain the shortened circadian period in aged rodents. Importantly, the model identifies the dosing schedules for maximizing the efficacy of anti-aging medications.

Use of Angiotensin-Converting Enzyme Inhibitors and Angiotensin II Receptor Blockers During the COVID-19 Pandemic: A Modeling Analysis.

Angiotensin-converting enzyme inhibitors (ACEi) and angiotensin II receptor blockers (ARB) are frequently prescribed for a range of diseases including hypertension, proteinuric chronic kidney disease, and heart failure. There is evidence indicating that these drugs upregulate ACE2, a key component of the renin-angiotensin system (RAS) and is found on the cells of a number of tissues, including the epithelial cells in the lungs. While ACE2 has a beneficial role in many diseases such as hypertension, diabetes, and cardiovascular disease, it also serves as a receptor for both SARS-CoV and SARS-CoV-2 via binding with the spike protein of the virus, thereby allowing it entry into host cells. Thus, it has been suggested that these therapies can theoretically increase the risk of SARS- CoV-2 infection and cause more severe COVID-19. Given the success of ACEi and ARBs in cardiovascular diseases, we seek to gain insights into the implications of these medications in the pathogenesis of COVID-19. To that end, we have developed a mathematical model that represents the RAS, binding of ACE2 with SARS-CoV-2 and the subsequent cell entry, and the host's acute inflammatory response. The model can simulate different levels of SARS-CoV-2 exposure, and represent the effect of commonly prescribed anti-hypertensive medications, ACEi and ARB, and predict tissue damage. Model simulations indicate that whether the extent of tissue damage may be exacerbated by ACEi or ARB treatment depends on a number of factors, including the level of existing inflammation, dosage, and the effect of the drugs on ACE2 protein abundance. The findings of this study can serve as the first step in the development of appropriate and more comprehensive guidelines for the prescription of ACEi and ARB in the current and future coronavirus pandemics.

Network centrality analysis of eye-gaze data in autism spectrum disorder.

Individuals suffering from autism spectrum disorder (ASD) exhibit impaired social communication, the manifestations of which include abnormal eye contact and gaze. In this study, we first seek to characterize the spatial and temporal attributes of this atypical eye gaze. To achieve that goal, we analyze and compare eye-tracking data of ASD and typical development (TD) children. A fixation time analysis indicates that ASD children exhibit a distinct gaze pattern when looking at faces, spending significantly more time at the mouth and less at the eyes, compared with TD children. Another goal of this study is to identify an analytic approach that can better reveal differences between the face scanning patterns of ASD and TD children. Face scanning involves transitioning from one area of interest (AOI) to another and is not taken into account by the traditional fixation time analysis. Instead, we apply four network analysis approaches that measure the "importance" of a given AOI: degree centrality, betweenness centrality, closeness centrality, and eigenvector centrality. Degree centrality and eignevector centrality yield statistically significant difference in the mouth and right eye, respectively, between the ASD and TD groups, whereas betweenness centrality reveals statistically significant between-group differences in four AOIs. Closeness centrality yields statistically meaningful differences in three AOIs, but those differences are negligible. Thus, our results suggest that betweenness centrality is the most effective network analysis approach in distinguishing the eye gaze patterns between ASD and TD children.