Time series (TS) modeling has come a long way from early statistical, mainly linear, approaches to the current trend in TS foundation models. With a lot of hype and industrial demand in this field, it is not always clear how much progress there really is. To advance TS forecasting and analysis to the next level, here we argue that the field needs a perspective. TS of observations from natural or engineered systems almost always originate from some underlying DS, and arguably access to its governing equations would yield theoretically optimal forecasts. This is the promise of , a class of ML/AI approaches that aim to infer of the underlying DS from data. But models based on DS principles offer other profound advantages: Beyond short-term forecasts, they enable to predict the of an observed system, which in many practical scenarios may be the more relevant quantities. DS theory furthermore provides domain-independent underlying TS generation, and thereby will inform us, e.g., about upper bounds on performance of TS model, generalization into unseen regimes as in tipping points, or potential control strategies. After reviewing some of the central concepts, methods, measures, and models in DS theory and DSR, we will discuss how insights from this field can advance TS modeling in crucial ways, enabling better forecasting with much lower computational and memory footprints. We conclude with a number of specific suggestions for translating insights from DSR into TS modeling.
Monoclonal antibodies targeting amyloid-β are the first approved disease-modifying treatment for Alzheimer’s disease. While amyloid-targeting therapies mitigate the progression of cognitive decline in early-stage Alzheimer’s disease, they are associated with amyloid-related imaging abnormalities (ARIA), an imaging phenomenon presenting as cerebral edema/effusion and/or hemorrhage. Redistribution of parenchymal amyloid-β to perivascular drainage pathways and direct antibody-amyloid interactions within the cerebral vasculature are considered key players in ARIA pathophysiology by promoting inflammation and vascular disruption, thus mirroring hallmarks of inflammatory cerebral amyloid angiopathy. Although ARIA is commonly regarded as an undesired side effect of amyloid-targeting therapies, its association with amyloid-β clearance from the brain opens up the possibility of an alternative interpretation as a physiological reaction to target engagement of anti-amyloid antibodies. Understanding risk factors that promote the occurrence of ARIA and its transformation from asymptomatic imaging phenomenon to its serious and severe form are of great importance to clinical practice. ARIA risk and severity are influenced by apolipoprotein E4 status, microvascular damage, and cerebral amyloid angiopathy, but may be further modulated by antibody binding preferences and comorbidities such as arterial hypertension and ischemic strokes. Identifying individual risk profiles based on deeper insights into pathophysiological pathways may improve patient safety and lead to personalized treatment concepts in Alzheimer’s disease. In this review, we provide a comprehensive summary of ARIA pathophysiology, highlight important risk factors and discuss their relevance in clinical risk management.
Psilocybin is studied as innovative medication in anxiety, substance abuse and treatment-resistant depression. Animal studies show that psychedelics promote neuronal plasticity by strengthening synaptic responses and protein synthesis. However, the exact molecular and cellular changes induced by psilocybin in the human brain are not known. Here, we treated human cortical neurons derived from induced pluripotent stem cells with the 5-HT2A receptor agonist psilocin - the psychoactive metabolite of psilocybin. We analyzed how exposure to psilocin affects gene expression, neuronal morphology, synaptic markers and neuronal function. Psilocin provoked a 5-HT2A-R-mediated augmentation of BDNF abundance. Transcriptomic profiling identified gene expression signatures priming neurons to neuroplasticity. On a morphological level, psilocin induced enhanced neuronal complexity and increased expression of synaptic proteins, in particular in the postsynaptic compartment. Consistently, we observed an increased excitability and enhanced synaptic network activity in neurons treated with psilocin. In conclusion, exposure of human neurons to psilocin might induce a state of enhanced neuronal plasticity, which could explain why psilocin is beneficial in the treatment of neuropsychiatric disorders where synaptic dysfunctions are discussed.
Glioblastomas are incurable primary brain tumors that depend on neural-like cellular processes, tumor microtubes (TM), to invade the brain. TMs also interconnect single tumor cells to a communicating multicellular network that resists current therapies. In this study, we developed a combined, comprehensive in vitro/in vivo anti-TM drug screening approach, including machine learning-based analysis tools. Two protein kinase C (PKC) modulators robustly inhibited TM formation and pacemaker tumor cell-driven, TM-mediated glioblastoma cell network communication. As TM-unconnected tumor cells exhibited increased sensitivity to cytotoxic therapy, the PKC activator TPPB was combined with radiotherapy, and long-term intravital two-photon microscopy paired with spatially resolved multiomics revealed anti-TM and antitumor effects. TPPB treatment also decreased the expression of tweety family member 1 (TTYH1), a key driver of invasive TMs. Our study establishes a novel screening pipeline for anti-TM drug development, identifies a TM master regulator pathway, and supports the approach of TM targeting for efficient brain tumor therapies. SIGNIFICANCE:Cancers can hijack neural properties to grow, disseminate, and to resist therapies, but effective drug development pipelines against these features are missing. Here, we establish a compound screening approach that allowed the identification of PKC modulators that target cancer cell-intrinsic neurodevelopmental mechanisms, suggesting a new class of neuroscience-instructed cancer therapeutics.
In dynamical systems reconstruction (DSR) we aim to recover the dynamical system (DS) underlying observed time series. Specifically, we aim to learn a generative surrogate model which approximates the underlying, data-generating DS, and recreates its long-term properties (`climate statistics'). In scientific and medical areas, in particular, these models need to be mechanistically tractable -- through their mathematical analysis we would like to obtain insight into the recovered system's workings. Piecewise-linear (PL), ReLU-based RNNs (PLRNNs) have a strong track-record in this regard, representing SOTA DSR models while allowing mathematical insight by virtue of their PL design. However, all current PLRNN variants are . This is in disaccord with the assumed continuous-time nature of most physical and biological processes, and makes it hard to accommodate data arriving at temporal intervals. Neural ODEs are one solution, but they do not reach the DSR performance of PLRNNs and often lack their tractability. Here we develop theory for PLRNNs (cPLRNNs): We present a novel algorithm for training and simulating such models, bypassing numerical integration by efficiently exploiting their PL structure. We further demonstrate how important topological objects like equilibria or limit cycles can be determined semi-analytically in trained models. We compare cPLRNNs to both their discrete-time cousins as well as Neural ODEs on DSR benchmarks, including systems with discontinuities which come with hard thresholds.
