T he scientific, academic, medical and data science communities have come together in the face of the COVID-19 pandemic crisis to rapidly assess novel paradigms in artificial intelligence (AI) that are rapid and secure, and potentially incentivize data sharing and model training and testing without the usual privacy and data ownership hurdles of conventional collaborations 1,2 . Healthcare providers, researchers and industry have pivoted their focus to address unmet and critical clinical needs created by the crisis, with remarkable results [3][4][5][6][7][8][9] . Clinical trial recruitment has been expedited and facilitated by national regulatory bodies and an international cooperative spirit 10-12 . The data analytics and AI disciplines have always fostered open

Background Machine learning models require large datasets that may be siloed across different health care institutions. Machine learning studies that focus on COVID-19 have been limited to single-hospital data, which limits model generalizability. Objective We aimed to use federated learning, a machine learning technique that avoids locally aggregating raw clinical data across multiple institutions, to predict mortality in hospitalized patients with COVID-19 within 7 days. Methods Patient data were collected from the electronic health records of 5 hospitals within the Mount Sinai Health System. Logistic regression with L1 regularization/least absolute shrinkage and selection operator (LASSO) and multilayer perceptron (MLP) models were trained by using local data at each site. We developed a pooled model with combined data from all 5 sites, and a federated model that only shared parameters with a central aggregator. Results The LASSOfederated model outperformed the LASSOlocal model at 3 hospitals, and the MLPfederated model performed better than the MLPlocal model at all 5 hospitals, as determined by the area under the receiver operating characteristic curve. The LASSOpooled model outperformed the LASSOfederated model at all hospitals, and the MLPfederated model outperformed the MLPpooled model at 2 hospitals. Conclusions The federated learning of COVID-19 electronic health record data shows promise in developing robust predictive models without compromising patient privacy.

Mitochondrial respiratory chain (RC) disease therapies directed at intra-mitochondrial pathology are largely ineffective. Recognizing that RC dysfunction invokes pronounced extra-mitochondrial transcriptional adaptations, particularly involving dysregulated translation, we hypothesized that translational dysregulation is itself contributing to the pathophysiology of RC disease. Here, we investigated the activities, and effects from direct inhibition, of a central translational regulator (mTORC1) and its downstream biological processes in diverse genetic and pharmacological models of RC disease. Our data identify novel mechanisms underlying the cellular pathogenesis of RC dysfunction, including the combined induction of proteotoxic stress, the ER stress response and autophagy. mTORC1 inhibition with rapamycin partially ameliorated renal disease in B6.Pdss2(kd/kd) mice with complexes I-III/II-III deficiencies, improved viability and mitochondrial physiology in gas-1(fc21) nematodes with complex I deficiency, and rescued viability across a variety of RC-inhibited human cells. Even more effective was probucol, a PPAR-activating anti-lipid drug that we show also inhibits mTORC1. However, directly inhibiting mTORC1-regulated downstream activities yielded the most pronounced and sustained benefit. Partial inhibition of translation by cycloheximide, or of autophagy by lithium chloride, rescued viability, preserved cellular respiratory capacity and induced mitochondrial translation and biogenesis. Cycloheximide also ameliorated proteotoxic stress via a uniquely selective reduction of cytosolic protein translation. RNAseq-based transcriptome profiling of treatment effects in gas-1(fc21) mutants provide further evidence that these therapies effectively restored altered translation and autophagy pathways toward that of wild-type animals. Overall, partially inhibiting cytosolic translation and autophagy offer novel treatment strategies to improve health across the diverse array of human diseases whose pathogenesis involves RC dysfunction.

Oxidative stress is a known contributing factor in mitochondrial respiratory chain (RC) disease pathogenesis. Yet, no efficient means exists to objectively evaluate the comparative therapeutic efficacy or toxicity of different antioxidant compounds empirically used in human RC disease. We postulated that pre-clinical comparative analysis of diverse antioxidant drugs having suggested utility in primary RC disease using animal and cellular models of RC dysfunction may improve understanding of their integrated effects and physiologic mechanisms, and enable prioritization of lead antioxidant molecules to pursue in human clinical trials. Here, lifespan effects of N-acetylcysteine (NAC), vitamin E, vitamin C, coenzyme Q10 (CoQ10), mitochondrial-targeted CoQ10 (MS010), lipoate, and orotate were evaluated as the primary outcome in a well-established, short-lived C. elegans gas-1(fc21) animal model of RC complex I disease. Healthspan effects were interrogated to assess potential reversal of their globally disrupted in vivo mitochondrial physiology, transcriptome profiles, and intermediary metabolic flux. NAC or vitamin E fully rescued, and coenzyme Q, lipoic acid, orotic acid, and vitamin C partially rescued gas-1(fc21) lifespan toward that of wild-type N2 Bristol worms. MS010 and CoQ10 largely reversed biochemical pathway expression changes in gas-1(fc21) worms. While nearly all drugs normalized the upregulated expression of the "cellular antioxidant pathway", they failed to rescue the mutant worms' increased in vivo mitochondrial oxidant burden. NAC and vitamin E therapeutic efficacy were validated in human fibroblast and/or zebrafish complex I disease models. Remarkably, rotenone-induced zebrafish brain death was preventable partially with NAC and fully with vitamin E. Overall, these pre-clinical model animal data demonstrate that several classical antioxidant drugs do yield significant benefit on viability and survival in primary mitochondrial disease, where their major therapeutic benefit appears to result from targeting global cellular, rather than intramitochondria-specific, oxidative stress. Clinical trials are needed to evaluate whether the two antioxidants, NAC and vitamin E, that show greatest efficacy in translational model animals significantly improve the survival, function, and feeling of human subjects with primary mitochondrial RC disease.

Mitochondrial respiratory chain (RC) diseases are highly morbid multi-systemic conditions for which few effective therapies exist. Given the essential role of sirtuin and PPAR signaling in mediating both mitochondrial physiology and the cellular response to metabolic stress in RC complex I (CI) disease, we postulated that drugs that alter these signaling pathways either directly (resveratrol for sirtuin, rosiglitazone for PPARγ, fenofibrate for PPARα), or indirectly by increasing NAD+ availability (nicotinic acid), might offer effective treatment strategies for primary RC disease. Integrated effects of targeting these cellular signaling pathways on animal lifespan and multi-dimensional in vivo parameters were studied in gas-1(fc21) relative to wild-type (N2 Bristol) worms. Specifically, animal lifespan, transcriptome profiles, mitochondrial oxidant burden, mitochondrial membrane potential, mitochondrial content, amino acid profiles, stable isotope-based intermediary metabolic flux, and total nematode NADH and NAD+ concentrations were compared. Shortened gas-1(fc21) mutant lifespan was rescued with either resveratrol or nicotinic acid, regardless of whether treatments were begun at the early larval stage or in young adulthood. Rosiglitazone administration beginning in young adult stage animals also rescued lifespan. All drug treatments reversed the most significant transcriptome alterations at the biochemical pathway level relative to untreated gas-1(fc21) animals. Interestingly, increased mitochondrial oxidant burden in gas-1(fc21) was reduced with nicotinic acid but exacerbated significantly by resveratrol and modestly by fenofibrate, with little change by rosiglitazone treatment. In contrast, the reduced mitochondrial membrane potential of mutant worms was further decreased by nicotinic acid but restored by either resveratrol, rosiglitazone, or fenofibrate. Using a novel HPLC assay, we discovered that gas-1(fc21) worms have significant deficiencies of NAD+ and NADH. Whereas resveratrol restored concentrations of both metabolites, nicotinic acid only restored NADH. Characteristic branched chain amino acid elevations in gas-1(fc21) animals were normalized completely by nicotinic acid and largely by resveratrol, but not by either rosiglitazone or fenofibrate. We developed a visualization system to enable objective integration of these multi-faceted physiologic endpoints, an approach that will likely be useful to apply in future drug treatment studies in human patients with mitochondrial disease. Overall, these data demonstrate that direct or indirect pharmacologic restoration of altered sirtuin and PPAR signaling can yield significant health and longevity benefits, although by divergent bioenergetic mechanism(s), in a nematode model of mitochondrial RC complex I disease. Thus, these animal model studies introduce important, integrated insights that may ultimately yield rational treatment strategies for human RC disease.