# Neural network analogies for immunological memory

Traditionally, the scientific consensus regarding host defense relied on a rigid dichotomy: an innate immune system that provided rapid, generic, and amnesic responses, and an adaptive immune system uniquely capable of immunological memory and antigen specificity. Over the last decade, this paradigm has undergone a profound transformation. Extensive research reveals that innate immune cells possess a highly functional, long-lasting memory mediated by epigenetic and metabolic reprogramming, challenging the assumption that only adaptive immunity can build memory [cite: 1, 2]. Concurrently, the discovery of dense, bidirectional communication networks between the central nervous system (CNS) and the immune system has collapsed the anatomical and conceptual boundaries that once separated them [cite: 3, 4, 5]. 

This synthesis of disciplines has spawned neuroscience-adjacent frameworks for understanding immune function. Researchers increasingly apply connectionist models, artificial neural network architectures, and Hebbian learning theories to explain how the immune system processes information, stores memory, and executes recall [cite: 6, 7, 8]. The immune system does not merely fight pathogens; it learns from them, adapts its baseline circuitry, and retains an environmental imprint that shapes future responses across the lifespan of the organism.

## Foundational Concepts of Adaptive Immune Memory

The architecture of immunological memory is stratified across distinct cellular lineages and compartments. The adaptive branch relies on somatic recombination to generate highly specific cellular receptors, selecting and refining these specificities over time to provide a tailored defense against previously encountered antigens.

### Specificity and Plasticity in T Lymphocytes

Adaptive immunological memory is orchestrated primarily by T and B lymphocytes. Following an initial encounter with a pathogen, naive T cells undergo rapid clonal expansion. Once the pathogen is cleared, the vast majority of these short-lived effector cells undergo apoptosis, leaving behind a resilient pool of memory cells that can be rapidly reactivated upon secondary exposure [cite: 2, 9]. Within the T cell compartment, memory is highly heterogeneous and defined by the expression of distinct surface markers, transcription factors, and migratory capabilities, which reflect the specific cellular state and tissue localization.

Central memory T cells (TCM) circulate through secondary lymphoid organs and are characterized by the expression of homing receptors such as CD62L, CD45RO, and CCR7, alongside the costimulatory molecule CD27 [cite: 9, 10, 11]. Genetic lineage tracing utilizing probabilistic labeling—such as the DivisionRecorder system—indicates that the TCM pool consists of cells with a high replicative history due to strong proliferative activity during the acute phase and selective proliferation after pathogen clearance [cite: 12]. Despite this history of division, a specific subset of TCM displays enriched expression of stem-cell-associated genes (such as *Tcf7*, *Ccr7*, and *Myb*) and exists in a quiescent state, endowing them with a high level of multipotency and the superior ability to elicit a proliferative recall response [cite: 12, 13]. This quiescent subset contrasts with an effector-like TCM subset that upregulates genes such as *Tbx21* and *Lgals1* [cite: 12].

Effector memory T cells (TEM), conversely, lack CCR7 and CD62L, allowing them to circulate freely through peripheral tissues. They exhibit markers associated with terminal differentiation, such as KLRG1 and CX3CR1, and are primed for immediate, antigen-specific cytokine production upon restimulation [cite: 9, 12]. 

The discovery of tissue-resident memory T (TRM) cells fundamentally altered the understanding of peripheral memory storage. First definitively characterized following herpes simplex virus infections, TRM cells are permanently lodged in non-lymphoid tissues—such as the skin, lungs, intestines, and reproductive tract—and do not recirculate to the blood [cite: 10, 13]. Their permanent retention is mediated by the continuous high-level expression of CD69. CD69 physically binds to and degrades the sphingosine-1-phosphate receptor 1 (S1PR1) at the cell surface, preventing the lymphocyte from following S1P gradients back into the systemic circulation [cite: 10]. In epithelial tissues, TRM cells additionally express CD103, an integrin that physically tethers the cells to the epithelial layer by interacting with E-cadherin. This specific molecular marker profile is activated by local transforming growth factor-beta (TGF-β) signaling, which is abundant in the epithelial microenvironment [cite: 10].

### B Cell Affinity Maturation and Germinal Center Dynamics

Humoral adaptive memory relies on the generation of memory B cells and long-lived plasma cells, a process largely restricted to the germinal centers (GC) of secondary lymphoid organs. During affinity maturation, light-zone GC B cells interact intimately with follicular dendritic cells (FDCs) and compete for limited survival signals provided by T follicular helper (Tfh) cells. This highly selective environment forces B cells to undergo somatic hypermutation (SHM) and class-switch recombination (CSR) to incrementally refine their B cell receptor (BCR) specificity [cite: 14, 15, 16]. 

The highest-affinity B cells are selected to exit the cell cycle and differentiate into durable memory populations. This transition requires a profound epigenetic reorganization of the cell, which halts transient *MYC* gene expression and actively induces the expression of the *PRDM1* gene (encoding Blimp-1), a master regulator of plasma cell differentiation [cite: 14, 16]. Human plasma cell precursors are characterized phenotypically by the loss of IL-4/STAT6 signaling responsiveness and the total absence of CD23 expression, differentiating them from the FDCs and uncommitted B cells that populate the light zone [cite: 14].

| Memory Cell Lineage | Primary Subsets | Key Molecular Markers | Localization and Functional Role |
| :--- | :--- | :--- | :--- |
| **T Lymphocytes** | Central Memory (TCM) | CD45RO+, CD62L+, CCR7+, CD27+ | Lymphoid organs; high proliferative capacity, stem-like multipotency, quiescent baseline [cite: 9, 10, 11]. |
| | Effector Memory (TEM) | CD45RO+, CD62L-, CCR7-, KLRG1+, CX3CR1+ | Peripheral circulation; rapid effector cytokine production upon restimulation [cite: 9, 12]. |
| | Tissue-Resident (TRM) | CD69+, CD103+ (epithelial), S1PR1- | Non-lymphoid tissues; immobile sentinels providing immediate local defense [cite: 10]. |
| **B Lymphocytes** | Memory B Cells | CD27+, CD23- (post-GC) | Circulation and lymphoid organs; rapid recall response and secondary expansion [cite: 14]. |
| | Plasma Cells | CD138+, PRDM1 (Blimp-1)+ | Bone marrow; sustained production of high-affinity neutralizing antibodies [cite: 14]. |

## Innate Immune Memory: The Concept of Trained Immunity

While adaptive memory requires structural genetic changes in the form of locus rearrangement, the innate immune system utilizes a fundamentally different mechanism to store information. Innate immune memory—termed "trained immunity"—is defined by the long-term functional reprogramming of innate immune cells through metabolic rewiring and epigenetic modifications [cite: 1, 17, 18]. Following primary exposure to certain sterile or infectious stimuli, cells such as monocytes, macrophages, natural killer (NK) cells, and innate lymphoid cells (ILCs) return to a baseline homeostatic state. However, they retain lingering epigenetic marks that alter chromatin accessibility, allowing for a dramatically enhanced, albeit nonspecific, antimicrobial and proinflammatory response upon secondary challenge with heterologous stimuli [cite: 2, 19, 20].

This process establishes a new physiological baseline. Cultured human monocytes subjected to training stimuli exhibit an altered cell surface phenotype long after the initial stimulus is removed, marked by increased expression of HLA-DR, CD14, CD40, and CD80, and a reduction in CD83 expression [cite: 21, 22, 23]. Single-cell RNA sequencing reveals that trained immunity is heterogeneous, differentiating monocytes into specific transcriptional clusters, including populations of reparative CX3CR1+ macrophages and non-classical CD14+CD16+ monocytes [cite: 22, 24, 25]. This memory phenomenon is evolutionarily conserved, operating in plants, invertebrates, and mammalian systems, highlighting its fundamental role in host survival prior to the evolutionary development of adaptive lymphocytes [cite: 19, 20, 26].

### Metabolic Rewiring and Energy Substrates

The induction and maintenance of trained immunity rely on an intricate, continuous feedback loop between cellular metabolism and the epigenome. The activation of specific metabolic cascades dictates the availability of substrates required by the enzymes that remodel chromatin, thereby linking the energy state of the cell directly to its transcriptional memory. 

The primary induction of trained immunity is characterized by a profound shift in cellular energy metabolism. When naive monocytes are exposed to prototypical training stimuli like β-glucan (a fungal cell wall component derived from *Candida albicans*) or the Bacillus Calmette–Guérin (BCG) vaccine, they rapidly upregulate aerobic glycolysis via the Akt/mTOR/HIF-1α signaling pathway [cite: 23, 27, 28]. This "Warburg-like" effect provides the rapid energy required for heightened inflammatory responses. Hexokinase 2 (HK2) recruitment to the mitochondria is a critical step in facilitating this robust glycolytic switch [cite: 27].

Concurrently, the tricarboxylic acid (TCA) cycle undergoes targeted fragmentation. This disruption leads to the intracellular accumulation of specific intermediary metabolites, most notably succinate and fumarate [cite: 27, 28, 29]. Succinate actively promotes inflammation by stabilizing HIF-1α, driving further transcription of glycolytic enzymes and the proinflammatory cytokine IL-1β. The mevalonate pathway and cholesterol synthesis are also upregulated, generating lipid intermediates like desmosterol that sustain the trained phenotype and activate signaling networks like the liver X receptor (LXR) [cite: 27, 28]. 

While glycolysis is the primary driver of proinflammatory trained immunity, oxidative phosphorylation (OxPHOS) is not completely abandoned. Monocytes trained with low doses of β-glucan or BCG show simultaneous enhancement of oxygen consumption and OxPHOS. This is mediated by Set7 methyltransferase, which modifies histones to upregulate genes associated with the TCA cycle. Pharmacological inhibition of ATP synthetase abolishes the trained phenotype in these contexts, indicating a complex metabolic requirement that extends beyond simple glycolysis [cite: 29]. Conversely, innate immune tolerance—a state of dampened responsiveness—is associated with suppressed glycolysis, reliance on OxPHOS, and the accumulation of metabolites like itaconate and α-ketoglutarate (αKG), which promote alternative anti-inflammatory phenotypes [cite: 18, 27, 28].

### Epigenetic Remodeling and Chromatin Accessibility

The cytosolic and mitochondrial metabolites generated during the induction phase act directly as essential co-factors or inhibitors for nuclear epigenetic enzymes. The physical translocation of substrates like acetyl-CoA and fumarate into the nucleus results in a profound remodeling of chromatin structure, converting transcriptionally repressive heterochromatin into accessible euchromatin at specific gene loci related to inflammation and metabolism [cite: 19, 28]. 

The primary epigenetic signatures of trained immunity include the broad accumulation of histone 3 lysine 4 trimethylation (H3K4me3) at gene promoters, which typically activates transcription, and histone 3 lysine 27 acetylation (H3K27ac) at distal enhancers [cite: 21, 23, 27, 28]. Fumarate, for instance, drives trained immunity by actively inhibiting KDM5 histone demethylases, thereby preserving existing H3K4me3 marks [cite: 27]. Additionally, long-term histone lactylation, catalyzed by the lactyltransferase p300, provides a direct mechanism connecting elevated glycolysis (and the resulting lactate production) with physical epigenetic rewiring [cite: 28]. 

Other enzymatic regulators strictly dictate the magnitude and persistence of the trained state. The G9a histone methyltransferase lays down the repressive H3K9me3 mark to inhibit responses, while Sirtuin 1, a histone deacetylase, can enhance the effects of BCG-induced memory [cite: 28, 29]. The baseline chromatin accessibility of an individual's myeloid cells can effectively predict whether they will be a robust or poor responder to trained immunity induction vaccines [cite: 28].

### Central Induction and Hematopoietic Stem Cell Reprogramming

A central paradox in the study of trained immunity has been reconciling the brief lifespan of circulating innate immune cells—often living only hours to days in the periphery—with longitudinal clinical data demonstrating that trained immunity persists for months to years in humans [cite: 30, 31, 32].

The longevity of trained innate immunity is achieved through a mechanism known as central induction. Training stimuli do not merely act on mature peripheral cells; they act systemically, penetrating the bone marrow microenvironment to fundamentally alter long-lived, self-renewing hematopoietic stem and progenitor cells (HSPCs) [cite: 19, 30, 31, 33]. Exposure to BCG, β-glucan, or systemic inflammation from a Western diet induces lasting epigenetic and transcriptional changes in specific subsets of HSPCs, most notably short-term hematopoietic stem cells (ST-HSCs) and multipotent progenitors (MPPs), marked by Lin−Sca1+c-Kit+CD34−CD150+CD48−Flt3− expression profiles [cite: 30, 31]. 

These reprogrammed progenitors exhibit increased chromatin accessibility at innate immunity genes, upregulate transcription factors like C/EBPβ, and bias their long-term differentiation toward the myeloid lineage [cite: 19, 29, 31]. Consequently, they produce a continuous lineage of functionally adapted daughter monocytes and macrophages. These progeny cells enter the peripheral circulation already possessing the trained epigenetic blueprint, capable of executing an enhanced inflammatory response despite having never directly encountered the original training stimulus [cite: 26, 32, 33].

| Mechanism Category | Pro-Inflammatory Trained Immunity | Innate Immune Tolerance |
| :--- | :--- | :--- |
| **Primary Inducers** | BCG, β-glucan, oxidized LDL, *Candida albicans* [cite: 17, 31, 34]. | High-dose LPS, specific helminth products, itaconate [cite: 18, 27, 35]. |
| **Metabolic Profile** | Upregulated aerobic glycolysis (Warburg effect), active Akt/mTOR/HIF-1α [cite: 27, 28]. | Suppressed glycolysis, reliance on oxidative phosphorylation (OxPHOS) [cite: 18, 27]. |
| **Key Metabolites** | Accumulation of succinate, fumarate, acetyl-CoA, desmosterol [cite: 27, 28, 29]. | Accumulation of itaconate, α-ketoglutarate (αKG) [cite: 27, 28]. |
| **Epigenetic Signatures** | Increased H3K4me3 (promoters), H3K27ac (enhancers), histone lactylation [cite: 21, 27, 28]. | Preserved repressive marks (H3K9me3 via G9a activity), loss of H3K27ac [cite: 27, 28]. |
| **Cellular Outcome** | Enhanced production of TNF-α, IL-6, IL-1β upon secondary stimulation [cite: 23, 29]. | Dampened cytokine response, shift toward M2 repair/regulatory phenotype [cite: 27, 36]. |

## Population Baselines and Environmental Imprinting

The functional outcome of trained immunity is highly dependent on the nature of the primary stimulus. This environment-dependent plasticity leads to shifting population immune baselines based on geographical and historical pathogen loads, directly influencing the prevalence of immune-mediated diseases.

While stimuli like β-glucan and BCG induce a classically pro-inflammatory trained phenotype, exposure to certain parasites induces a distinct, anti-inflammatory trained state. Chronic infection with helminths (intestinal worms) or exposure to their derived glycans reprograms host hematopoietic stem cells to generate peripheral monocytes and innate lymphoid cells (ILC2s) biased heavily toward type 2 immunity. This trained phenotype is characterized by the production of IL-5, IL-10, and IL-33, the expansion of Ly6Clow regulatory monocytes, and the facilitation of regulatory T cells [cite: 19, 35, 37]. 

This helminth-induced immune regulation forms the biological basis of the "hygiene hypothesis." In industrialized populations where endemic parasitic infections have been medically eradicated, the absence of this regulatory innate training creates an evolutionary mismatch. The population baseline shifts toward hyper-reactivity, directly correlating with increased regional incidences of autoimmune disorders, asthma, allergies, and inflammatory bowel disease [cite: 22, 35, 37].

Conversely, chronic infections like malaria (*Plasmodium falciparum*) establish a uniquely hyper-reactive baseline. The accumulation of hemozoin during infection drives persistent pro-inflammatory TLR2 signaling and widespread H3K4me3 enrichment in myeloid cells. While this trains the innate immune system, the *Plasmodium* parasite simultaneously secretes extracellular vesicles (EVs) that inhibit specific T-cell responses, creating a complex immunological state where innate hyper-inflammation occurs alongside adaptive immune evasion [cite: 19, 31, 35].

## Memory Erasure and Active Resolution

Immunological memory is not immutable. When innate immune memory persists maladaptively—such as the pro-inflammatory metabolic memory retained by pancreatic macrophages following high-fat diet (HFD) induced obesity—it contributes to chronic pathology. In models of obesity, even when subjects return to normal diets and systemic cytokine levels normalize, the innate immune response of myeloid cells retains the epigenetic marks of the HFD period, driving persistent beta-cell inflammation and the progression of Type 2 Diabetes [cite: 38]. Resolving these pathological states requires active memory erasure.

### Pharmacological and Epigenetic Erasure

The erasure of innate immune memory is an active, biochemically regulated process. Epigenetic modulators have demonstrated the ability to reset pathological metabolic memory. Vitamin C (VIC), acting as an essential reducing cofactor, enhances the activity of α-ketoglutarate-dependent dioxygenases (α-KGDD) and the ten-eleven translocation (TET) family of DNA hydroxylases. By facilitating widespread DNA and RNA demethylation, Vitamin C effectively erases pathological epigenetic memory in somatic and immune cells, reversing long-term chromatin alterations and returning cells to an untrained, homeostatic state [cite: 39].

In oncology, the concept of "mechanical memory" parallels trained immunity. Cancer cells exposed to stiff extracellular matrices (ECM) undergo epigenetic reprogramming—often mediated by the sustained nuclear localization of YAP and the upregulation of MicroRNA-21—that locks them into a rigid, metastatic phenotype. If the duration of mechanical priming is short, the cellular memory is reversible. However, prolonged exposure leads to irreversible epigenetic alterations that exacerbate fibrosis and metastasis, prompting ongoing research into pharmacological memory erasure concepts targeting these mechanotransduction pathways [cite: 40].

### Microglial Memory and Synaptic Pruning in the CNS

In the central nervous system, microglia (the resident innate macrophages of the brain) exhibit robust immune memory that profoundly influences neurological health. Following repeated peripheral inflammatory insults (such as low-dose LPS), microglia undergo long-term transcriptomic and morphological changes, entering states of either trained immunity or immune tolerance [cite: 41].

Trained microglia rely heavily on enhanced glycolysis and exhibit prolonged expression of epigenetic marks like H3K4me1 and H3K27ac. In neurodegenerative contexts like Alzheimer's disease, trained microglia exacerbate pathology by over-producing pro-inflammatory cytokines and aggressively phagocytosing neuronal synapses, accelerating cognitive decline [cite: 27, 41]. 

The active "forgetting" or erasure of memories in the brain is physically executed by these immune cells. Studies tracking the transition of memory engrams from a latent to a silent phase demonstrate that microglia actively eliminate synaptic components in the adult hippocampus. Depleting microglia or pharmacologically inhibiting their phagocytic capacity prevents the degradation of these memory engrams, effectively halting the forgetting of previously learned contextual fear memories [cite: 41, 42]. This indicates that active physical clearance mechanisms by the innate immune system are required to terminate the signaling loops that sustain persistent neural memory.

## Immune Imprinting and Original Antigenic Sin

While trained innate immunity explores the broad enhancement of host defense, the adaptive immune system faces a highly specific challenge: managing the evolutionary trade-off between the rapid recall of existing antibody memory and the flexibility required to adapt to rapidly mutating pathogens. This dynamic is governed by the phenomenon of immune imprinting, historically termed "original antigenic sin" (OAS).

### Evolutionary Trade-Offs in Adaptive Responses

Upon encountering a novel, mutated pathogen strain, the adaptive immune system relies heavily on its immunological history. Pre-existing memory B cells that recognize conserved epitopes across variant strains are activated far more rapidly than naive B cells can be recruited, vetted, and trained [cite: 43, 44]. This rapid activation results in a secondary response heavily dominated by a bimodal B cell receptor (BCR) repertoire biased toward high-affinity clones for the ancestral strain [cite: 44]. 

While this systemic bias ensures a swift and massive antibody response, it actively suppresses the generation of *de novo* naive responses targeting the newly mutated epitopes of the variant [cite: 43, 45, 46]. Mathematical modeling demonstrates that the generation of new, strain-specific memory is strictly constrained by a "homology threshold" or "blunting distance." Only when a challenge strain escapes existing immunity at nearly every epitope does the system overcome the kinetic dominance of cross-reactive memory cells and authorize the creation of a fundamentally new memory profile [cite: 43, 44].

This dynamic reveals a profound structural and evolutionary trade-off between the humoral and cellular branches of the adaptive system. Humoral memory (antibodies) targets fragile, three-dimensional conformational epitopes, providing sterilizing but highly strain-specific protection that is easily abrogated by slight structural mutations (antigenic drift) [cite: 44]. Conversely, cellular memory (CD4+ and CD8+ T cells) targets linear peptide sequences derived from intracellular processing, which are structurally constrained and highly conserved across variants. Thus, while antibodies provide sterilizing immunity that wanes rapidly against new variants, T cell memory provides broad, cross-reactive protection that prevents severe disease but does not prevent initial infection [cite: 44, 47, 48]. 

### Clinical Observations in Viral Pathogenesis

The clinical impacts of immune imprinting and evolutionary trade-offs are starkly visible in the trajectories of human respiratory viruses. Viral evolution is primarily driven by the selective pressure to evade pre-existing host immunity. In SARS-CoV-2 variants—such as Delta, Omicron (BA.1, BA.2), and XBB—mutations in the spike protein exhibit convergent evolution driven directly by immune imprinting from previous infections and vaccinations [cite: 49, 50, 51].

The virus navigates this landscape through distinct evolutionary pathways. An *enthalpic trade-off* occurs when mutations in the receptor-binding motif (RBM) successfully evade neutralizing antibodies but incur a severe cost to ACE2 receptor binding affinity. Alternatively, the virus employs an *entropic trade-off*, where mutations outside the binding motif shift the protein's overall conformational equilibrium, reducing the time the spike protein spends in the "open" state to hide from immune surveillance, without directly altering the binding interface itself [cite: 49, 50].

Longitudinal studies demonstrate that antibody levels wane rapidly, particularly against heavily mutated variants, as immune imprinting restricts the breadth of the active neutralizing antibody repertoire [cite: 45, 46, 52, 53]. However, T cell immunity maintains remarkable durability and cross-reactivity. SARS-CoV-2 specific memory T cells, alongside persistent effector functionalities in NK and cytotoxic T cells, remain robust for 8 to 12 months post-infection, largely unaffected by the antibody-evading mutations that define novel variants [cite: 47, 48, 54]. Interestingly, the T cell memory profile for SARS-CoV-2 (characterized by specific CD8+:CD4+ ratios and high IL-2/IL-6 production) differs distinctly from the lifelong memory profile established by repeated, lifelong exposures to the Influenza A virus, suggesting that pathogen chronicity and exposure frequency permanently sculpt the precise phenotype of the cellular memory compartment [cite: 55].

| Pathogen / Response Type | Primary Target Antigen | Durability & Efficacy | Impact of Immune Imprinting |
| :--- | :--- | :--- | :--- |
| **Humoral (Antibody) Memory** | 3D Conformational Epitopes (e.g., Spike RBM) [cite: 44]. | Fragile; provides sterilizing immunity but wanes rapidly [cite: 44, 52]. | High; rapid recall of ancestral antibodies suppresses *de novo* responses to new variants [cite: 43, 44]. |
| **Cellular (T-Cell) Memory** | Linear intracellular peptides (highly conserved) [cite: 44]. | Robust; lasts 8–12+ months; protects against severe disease [cite: 48, 54]. | Low; maintains broad cross-reactivity across highly mutated variant lineages [cite: 47, 48]. |
| **SARS-CoV-2 Evolution** | Spike Protein / ACE2 interface [cite: 50, 51]. | Rapid antigenic drift via enthalpic and entropic mutations [cite: 49, 50]. | Imprinting forces convergent evolution across sub-lineages due to narrowed antibody pools [cite: 51]. |

## The Neuro-Immune Axis

The nervous and immune systems do not operate as isolated, compartmentalized silos; they form a singular, integrated physiological ecosystem. Both systems execute complex sensory, computational, and effector functions to maintain whole-body homeostasis. They share a common chemical vocabulary of cytokines, neurotransmitters, and neuropeptides, enabling continuous, bidirectional communication that directly modulates immune memory, inflammation, and cellular function across distance [cite: 3, 4, 5, 56].

### Hypothalamic and Autonomic Regulation

The autonomic nervous system physically innervates primary and secondary lymphoid organs (such as the bone marrow, spleen, and lymph nodes), exerting direct, hardwired control over the immune response and the generation of memory [cite: 36, 57]. 

The systemic response to internal and external threats is regulated by the hypothalamic-pituitary-adrenal (HPA) axis. Psychological or physiological stress prompts the hypothalamus to secrete corticotropin-releasing hormone (CRH). CRH acts directly on endothelial cells and macrophages to upregulate adhesion molecules and stimulate pro-inflammatory cytokine release. Concurrently, CRH stimulates the pituitary gland to release adrenocorticotropic hormone (ACTH), driving the adrenal production of systemic glucocorticoids and norepinephrine (NE) [cite: 36, 57]. 

Norepinephrine acts as a powerful immunomodulator by binding directly to β2-adrenergic receptors (β2-ARs) highly expressed on macrophages and T cells. In innate myeloid cells, this adrenergic signaling suppresses pro-inflammatory cytokine production and actively shifts macrophages toward an alternative (M2) repair phenotype [cite: 36, 58]. In the adaptive compartment, sympathetic innervation suppresses Th1-mediated cellular responses while promoting Th2, Th17, and regulatory T cell development, thereby actively steering the nature of the developing adaptive memory [cite: 57]. 

Parasympathetic regulation is mediated primarily by the vagus nerve, which establishes a systemic "cholinergic anti-inflammatory pathway." Acetylcholine released by vagal efferents binds to α7-nicotinic acetylcholine receptors (nAChR) on innate cells residing in the spleen and gut, potently inhibiting the release of tumor necrosis factor (TNF) and damping systemic vascular inflammation [cite: 36, 57, 59]. Furthermore, a wide array of neuropeptides—including vasoactive intestinal peptide (VIP), Substance P, gamma-aminobutyric acid (GABA), and calcitonin gene-related peptide (CGRP)—are secreted by sensory nerves directly into tissue microenvironments. These peptides bind to specific receptors on innate lymphoid cells and antigen-presenting cells to fine-tune the local immune baseline and modulate the tumor microenvironment [cite: 57, 60].

## Connectionist Models of Immune Learning

Because the immune and nervous systems exhibit profound architectural similarities—both relying on decentralized, distributed networks of discrete cells to sense, learn, and remember complex environmental patterns—computational neuroscientists and immunologists increasingly utilize models from artificial intelligence and neural network theory to map immunological behavior [cite: 4, 6, 8].

### Hebbian Learning Analogies in Immunological Networks

Donald Hebb's seminal neuropsychological postulate—colloquially summarized as "neurons that fire together, wire together"—describes associative synaptic plasticity wherein persistent, correlated presynaptic and postsynaptic stimulation increases synaptic efficacy [cite: 61]. This foundational concept of unsupervised learning maps exceptionally well to the biological dynamics of immune network activation [cite: 7, 62, 63]. 

In immunology, the concept of the "immune synapse" emerged as a direct structural parallel to the neurological synapse. When an antigen-presenting cell (APC) interacts with a T cell or B cell, the physical contact zone involves the dynamic, targeted clustering of receptors, adhesion molecules, and signaling complexes [cite: 4, 8]. If the signaling is persistent and the threshold for activation is met, the biological "connection" between that specific antigen pattern and the lymphocyte is strengthened exponentially through rapid clonal expansion—a biological execution of a Hebbian weight update [cite: 8, 64].

Recent connectionist models of the immune system and epidemiology utilize Differential Hebbian Learning (DHL) rules, which account for the specific temporal sequence of events, mirroring spike-timing-dependent plasticity in cortical circuits [cite: 61, 65, 66]. In these conceptual models, positive reinforcement (repeated, successful antigen encounters) increases the "activation rate" of a specific immune pathway, solidifying long-term immunological memory. Conversely, negative reinforcement (anti-Hebbian learning) acts analogously to innate and adaptive immune tolerance. When repeated sub-threshold stimulation or regulatory signaling occurs, the network actively degrades connections, leading to targeted unresponsiveness to harmless environmental antigens, self-antigens, or commensal microbes [cite: 61, 64].

### Artificial Neural Networks Inspired by Immunology

The mathematical and conceptual symmetry between the systems is so robust that computer scientists now engineer artificial neural networks (ANNs) based directly on immune logic to overcome the fundamental biological implausibility and computational limitations of standard algorithms [cite: 67, 68]. 

Classical AI training methods, such as backpropagation, require global error signals, strict weight symmetry, and perfect non-local credit assignment—features entirely absent in biological neural and immune networks, where cells only have access to local information [cite: 67, 68]. Hebbian learning rules, contrastive Hebbian learning (CHL), and Hopfield associative memory networks offer decentralized, unsupervised alternatives that mimic biological reality [cite: 7, 69, 70]. In these systems, networks self-organize by mapping the statistical overlap of inputs, exactly as the immune system maps overlapping epitopes.

By modeling AI algorithms on the adaptive immune system's process of generating specific B cell antibodies to neutralize unpredictable antigens, researchers have developed robust, self-organizing computational networks. These immune-inspired models exhibit "hysteresis" (state-dependency at the single-unit level) and are highly effective at maintaining stability and defending against adversarial attacks in complex systems, such as power grid control networks. This mirrors the biological immune system's ability to maintain homeostasis and execute targeted responses against diverse, constantly evolving pathogenic threats without a centralized controller [cite: 6, 71]. 

Furthermore, applying synapse-type-specific competitive Hebbian learning—where excitatory and inhibitory "synapses" compete for separate resource pools—allows for the spontaneous development of stable, inhibition-balanced receptive fields, explaining how both cortical and immune networks avoid runaway positive feedback loops [cite: 72, 73].

| Network Characteristic | Central Nervous System (Biological & Artificial) | Immune System (Biological Network) |
| :--- | :--- | :--- |
| **Basic Processing Unit** | Neurons / Perceptrons [cite: 7, 62]. | Lymphocytes (T/B cells) and Myeloid Cells [cite: 8]. |
| **Communication Interface** | Neural Synapse (Neurotransmitters/Electrical) [cite: 62]. | Immune Synapse (Cytokines/Receptor engagement) [cite: 4, 8]. |
| **Learning Mechanism** | Hebbian Plasticity (Synaptic weight updates) [cite: 7, 61, 67]. | Clonal Selection & Epigenetic Reprogramming [cite: 8, 64]. |
| **Memory Storage** | Distributed associative memory in neural engrams [cite: 42, 69]. | Systemic distribution of expanded memory cell subsets [cite: 4, 8]. |
| **Forgetting / Erasure** | Synaptic pruning, depotentiation, active engram silencing [cite: 41, 42]. | Apoptosis of effector cells, active epigenetic demethylation [cite: 2, 39]. |
| **Tolerance/Suppression** | Anti-Hebbian learning, inhibitory interneuron feedback [cite: 61, 64]. | Regulatory T cells, macrophage M2 polarization, exhaustion [cite: 36, 57]. |

## Conclusion

The characterization of the immune system solely as a reactive, amnesic defense mechanism has been wholly superseded by a comprehensive understanding of it as a decentralized, cognitive network capable of complex computation and memory storage. Immunological memory spans a vast biological continuum. It ranges from the rapid, highly specific structural adaptations of B and T lymphocytes undergoing somatic hypermutation within germinal centers, to the broad, epigenetically encoded metabolic memories maintained by innate myeloid cells and central hematopoietic stem cells. 

This intricate physiological system is deeply enmeshed with the central nervous system, subjected to direct autonomic and hypothalamic regulation that fine-tunes the critical balance between memory formation, inflammation, and immune tolerance. By interpreting the immune system through the lens of connectionist models and Hebbian learning theories, researchers have illuminated the underlying algorithms that govern how the body learns from its environment, stores pathogenic data, and recalls information. This neuroscience-adjacent approach not only clarifies the pathogenesis of complex autoimmune, infectious, and neurodegenerative diseases but also paves the way for advanced immunotherapeutic interventions, active memory erasure protocols, and the next generation of biologically plausible artificial intelligence.

## Sources
1. [Frontiers in Immunology](https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2023.1147476/full)
2. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC7186935/)
3. [JCI Insight](http://insight.jci.org/articles/view/195866/files/pdf)
4. [FEMS Microbiology Reviews](https://academic.oup.com/femsre/article/doi/10.1093/femsre/fuaf036/8230974)
5. [PubMed](https://pubmed.ncbi.nlm.nih.gov/3796008/)
6. [Wikipedia](https://en.wikipedia.org/wiki/Hebbian_theory)
7. [The Decision Lab](https://thedecisionlab.com/reference-guide/neuroscience/hebbian-learning)
8. [Medium](https://medium.com/@reutdayan1/hebbian-learning-biologically-plausible-alternative-to-backpropagation-6ee0a24deb00)
9. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC6130884/)
10. [Frontiers in Immunology](https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2025.1666233/full)
11. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC12680377/)
12. [Physiological Reviews](https://journals.physiology.org/doi/full/10.1152/physrev.00031.2021)
13. [eLife](https://elifesciences.org/articles/106610)
14. [Frontiers in Immunology](https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2026.1777470/pdf)
15. [Frontiers in Immunology](https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2026.1822561/full)
16. [bioRxiv](https://www.biorxiv.org/content/10.64898/2026.01.30.702892v1.full-text)
17. [MDPI](https://www.mdpi.com/2072-6694/8/3/36)
18. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC7612726/)
19. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC9313577/)
20. [ResearchGate](https://www.researchgate.net/publication/356725627_Committed_Human_CD23-Negative_Light-Zone_Germinal_Center_B_Cells_Delineate_Transcriptional_Program_Supporting_Plasma_Cell_Differentiation)
21. [ACS Nano](https://pubs.acs.org/doi/10.1021/acsnano.5c04836)
22. [MDPI](https://www.mdpi.com/1999-4915/11/2/122)
23. [MDPI](https://mdpi-res.com/bookfiles/book/11885/Cellular_and_Molecular_Mechanisms_in_Immune_Regulation.pdf?v=1764301579)
24. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC12513719/)
25. [eLife](https://elifesciences.org/articles/108465)
26. [eLife](https://elifesciences.org/articles/105393)
27. [Frontiers in Immunology](https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2020.00308/full)
28. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC12570465/)
29. [Ovid](https://www.ovid.com/00004268-201204000-00002)
30. [AHA Journals](https://www.ahajournals.org/doi/10.1161/hypertensionaha.111.186833)
31. [Frontiers in Immunology](https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2025.1619338/full)
32. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC12703160/)
33. [Ovid](https://www.ovid.com/journals/scie/pdf/10.1126/science.ady9543~convergence-and-divergence-of-individual-immune-responses)
34. [MDPI](https://www.mdpi.com/2076-393X/13/4/424)
35. [PNAS](https://www.pnas.org/doi/10.1073/pnas.2400202121)
36. [JCI](https://www.jci.org/articles/view/163202)
37. [PubMed](https://pubmed.ncbi.nlm.nih.gov/3796008/)
38. [IIT](https://www.iit.edu/news/giving-neural-networks-immune-system)
39. [Stanford Encyclopedia of Philosophy](https://plato.stanford.edu/archives/fall2019/entries/connectionism/)
40. [Wikipedia](https://en.wikipedia.org/wiki/Connectionism)
41. [Frontiers in Immunology](https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2022.984678/full)
42. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC5087274/)
43. [eLife](https://elifesciences.org/articles/106597)
44. [Journal of Leukocyte Biology](https://academic.oup.com/jleukbio/advance-article/doi/10.1093/jleuko/qiag047/8644458)
45. [Frontiers in Immunology](https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2023.1147476/full)
46. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC9575643/)
47. [ResearchGate](https://www.researchgate.net/publication/391744937_A_Restarted_Life_Resetting_Microglia_Innate_Immune_Memory)
48. [World Journal of Gastroenterology](https://www.wjgnet.com/1948-9358/full/v17/i5/118754.htm)
49. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC12609669/)
50. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC11820684/)
51. [bioRxiv](https://www.biorxiv.org/content/10.64898/2026.04.12.717978v1.full)
52. [Frontiers in Immunology](https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2024.1412873/full)
53. [Virus Evolution](https://academic.oup.com/ve/article/11/1/veaf027/8159764)
54. [PNAS](https://www.pnas.org/doi/10.1073/pnas.2536956123)
55. [bioRxiv](https://www.biorxiv.org/content/10.64898/2026.02.26.708225v1.full.pdf)
56. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC12901377/)
57. [bioRxiv](https://www.biorxiv.org/content/10.1101/2025.05.09.653036v1)
58. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC11688642/)
59. [MDPI](https://www.mdpi.com/1422-0067/22/19/10684)
60. [Frontiers in Immunology](https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2022.872652/full)
61. [Esser-Kahn Lab](https://www.esser-kahnlab.org/trained-immunity)
62. [bioRxiv](https://www.biorxiv.org/content/10.1101/2022.03.11.483899v3)
63. [Wikipedia](https://en.wikipedia.org/wiki/Hebbian_theory)
64. [Emergent Mind](https://www.emergentmind.com/topics/hebbian-learning)
65. [The Decision Lab](https://thedecisionlab.com/reference-guide/neuroscience/hebbian-learning)
66. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC11194505/)
67. [World Journal of Gastroenterology](https://www.wjgnet.com/1948-9358/full/v17/i5/118754.htm)
68. [ResearchGate](https://www.researchgate.net/publication/391744937_A_Restarted_Life_Resetting_Microglia_Innate_Immune_Memory)
69. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC11820684/)
70. [bioRxiv](https://www.biorxiv.org/content/10.64898/2026.04.12.717978v1.full)
71. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC12609669/)
72. [Frontiers in Immunology](https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2022.872652/full)
73. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC5550561/)
74. [ResearchGate](https://www.researchgate.net/figure/Heterogeneous-trained-immunity-effect-in-terms-of-expression-of-marker-genes-among_fig3_358446744)
75. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC9259164/)
76. [bioRxiv](https://www.biorxiv.org/content/10.1101/2023.06.10.544264v1.full-text)
77. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC9143792/)
78. [News-Medical](https://www.news-medical.net/news/20210614/Immune-memory-responses-differ-between-SARS-CoV-2-and-influenza-A.aspx)
79. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC11832264/)
80. [MDPI](https://www.mdpi.com/2076-393X/10/5/696)
81. [ScienceOpen](https://www.scienceopen.com/document_file/a6c303c3-3beb-4224-9d4a-e7aa8b0a6270/PubMedCentral/a6c303c3-3beb-4224-9d4a-e7aa8b0a6270.pdf)
82. [Learn Medical Neuroscience](https://www.learnmedicalneuroscience.nl/sfn-2025-michal-schwarz-the-brain-immune-ecosystem/)
83. [Gordon Research Conferences](https://www.grc.org/neuroimmune-communication-in-health-and-disease-conference/2025/)
84. [NCBI PMC](https://pmc.ncbi.nlm.nih.gov/articles/PMC3314485/)
85. [News-Medical](https://www.news-medical.net/news/20250213/How-immune-cells-and-neurons-secretly-shape-brain-health-behavior-and-disease.aspx)
86. [PubMed](https://pubmed.ncbi.nlm.nih.gov/40640567/)
87. [arXiv](https://arxiv.org/pdf/2508.21053)
88. [bioRxiv](https://www.biorxiv.org/content/10.1101/2025.01.16.633317v1.full-text)
89. [Annual Reviews](https://www.annualreviews.org/content/journals/10.1146/annurev-physchem-083122-010308?crawler=true&mimetype=application/pdf)
90. [PNAS](https://www.pnas.org/doi/10.1073/pnas.2309876121)
91. [ScienceOpen](https://www.scienceopen.com/hosted-document?doi=10.15212/bioi-2025-0199)

**Sources:**
1. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGa7rI7PBLtFlRe2fUfAuGJNbZthFSOdcsVBJO_lHsLxTO_gd9lN4sLrBGINa6L3bS5RLUDDwPn1Odjk298EzMeUSyMmpUeax0IQSoWRizUSJLJ0_R0i2budi9nQJ22fdLmqg5gvM4=)
2. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFBz8twACd0ooTe4qfzCiVxPP7ig1cq9Q3NRNvxPU0br_TibrqbqZfw2FVTjK8Q52jdJdC6GUWBN33FnIBOLJe_4eMxOH2fQQweO-EQKcLiebsoEOTC8EXW8ekhL5MmMdUkT_6QYOs=)
3. [learnmedicalneuroscience.nl](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGaCePa55jpkU3KHtBxRX54lUmLpvneI-3yT3ANx1XIaSE8DNKuVQ0goqTK_xeUuLXnlXHH0vjHWMvYpFk5wJbaCGDIkLtS1Vg8_ScDzan3wVr9CHaNkI3_oB5jLrFddjfmGrc66T-2WMvTuQo8Lc-JR10p3gA3ngT8FVdk4xTQpabMT3NXshLU3odzhJK2k9o=)
4. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHcyfSEHjFeePllgZKNcciaWBZ8SN_1ggr7Tab2V11oJpdxKU5MPcOb1ypC_oNg215_Rx9VXKxF0VIPUMGB1R1gu7q5jpAcjlsA4n6fAEpDDh5oXGGRThZLvlGCiXogWYlUb0pHUuc=)
5. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGVjhrpFNnvRgXyXtnl6RGDlBqfJwYX6UakQs_VcTwa6cghORi3wIN0F1tImvzxM0QDR8gEXoOqcyN_hVvdiS1tJxcRiF90W0qOBnDHlgS7eKJqqDAFdtSSbzUAwddl)
6. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQF8W7jJ6Z5OKdrJws48P1ksv09HF39oWCOYkrH94fApmhqtZy_H4mt-ZgRKAJaYyOPFlzZOT2xDGuocWWtra9YfSQ7x_xrM3XHJ64vGkGOU22ikPza4LasFcxWyaLQ=)
7. [stanford.edu](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEhcRmAzotyoow7obEq24-NQXv9hwM7DnV1tj6gKwzOHdsh2pRtEvFHAfJv3ZBVFod6IJUQWjRpBYrXvb0NGnntKJiFjZIKcl_lBvx9QntoyKvPp_bPT6jm-XjgYHdyCUyRXI6cdRKgi_Wz6em7W7rU_qa54WJTLHY=)
8. [frontiersin.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGuHGLSNJewWpVhYxXixBlEQS6Yw4P2ebfuiMIyG-Ds6p9mzlTxPm0G8FgNQ65btGvSpPAXYZS5Hn8NBq_AfYap1x5_FpzVyHPTGEUGgoTWczI7Lz25jkxcWsoUuSXhSAnu3u2cGc-j5alRHWapqJUvslQjzSVfgtnJS8adkw-_fBA-pH5FqC4v4YU7Vw==)
9. [mdpi.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFUrADFaA9PyZObXme_c8VkULfZg2_ew3t7DXBgegf_opX-yyGfqpwmA2vrRq_w8l69mJ9VJQ8jOiirmKRx9zg0ngifOQn9WU1zNoqV-oEg0CaYez7Nd8fJZmE=)
10. [frontiersin.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHnKKEEX_OpVfOqiTxC7hTe_P-vWhf-WsCnTcYTrFiaFhf-lm3wuJGAMETJZ-D_N5y3aEEtIVPewmENtz0Vdan8cxuuBIJZ-YjCfPSBOslTLsY2e22azzDPolWZWja9O9fmC_FRqPQj6hm8SB1BkKy13fuGv24QgmdSXKnVORxozqK0qZuqyD8MZF-w2xw=)
11. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEG_mZgV_Bis6RlYuLpWm6MvWCWye8mLOkmiYaqyTWs7Ees7KZ4Q-20S3fMkRFcxQqPykbn8TMsYESAA0BhNGBSyBONaHVfS4qEGmEBR_A97WNNQs-JOyNpFJhXwS8Ehua69CKbuy0=)
12. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHpwqdNrIvAQEt_5bBRBwlfvF-MS_QOusG39vQlJ8r3PxBZMEzYc2z1FN-dO8cwK758k3AGRAdKJHQUT-27-Z5eyEYhYuOANC_fPP5ml2hNmbIr3L2XU6u9QB-00iPFDffZDZDCcx0=)
13. [biorxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHiROQNcNjezPm9YESAuF9P3qyTHmZQasXVchFZg5kN7MLYgUzV6Kax2gv9qm_QJovAPFZymF84Bg3S5m7hK07epH5xO7hevi6j8HTeYqD9z9vULdc5fMVqEfs9Z2gdyueK2UKx9W8OvB7gyo49aepY8h8nNvTLsl6tF_E=)
14. [researchgate.net](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFgEjtNn7kbfKYzM6GJSBb6c1MvFxBvvMUpaOh3gZFMeJ9nz4Hu3jGW1kuTdeY7iLkivOR_6qE0rvJwWk9M6z8tQwvwm8XgZHikeXMVp2Ut6M9L12CCYhVslTx9vr8J8BozO-FgUzaGiyrv3DjNcjzEbnMutGGQbHb3TrhaKbShlNHSRanYfOrPu-vCqjqW5HCC_lPqhZ7zLxKToK9UuE6Bd7HU7d4U0NsLFcTGeXFQf-4QEZStv7gPvUA_NedkxlIhK3laKnKvQDfnkSMOKoDLdU5TgNsk9jVJTmGd_LoKU7Zku1x_8g3ma6mAHN3zKTkLkA==)
15. [mdpi.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGZYMgtyUILLcrnlGHiNUGJ5RUhNrUD6FnWRK3IVtFmdWC9Dqww8PL9sPm-mukWjDWixRVcdc8PGnQK4_NqoqPpfgEk16ojd5OYd8GH_ijruja-6UAaca_QlJq5ew==)
16. [mdpi-res.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFKsKjxYxg3mVPjtNCJrwFDQVDbVlMHppl14VHXAZtCMuQffaPm96fijxvFFTe4vL0euX2E165rRX-iB0Q6coerxlAAcUqyhXksf5umPcbcumEcypazp1cscIiBtbOk8RJkuWxDtjTYTG2OoGYMq-HnIjOEadPVgrFUY-Nw71EBxGkNf2NMprDMcm7dkZQ7ZHV6MVCkPzWM5NA5qHMN6VXkI2F-lfKE)
17. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFm17EgIig2lch_EZOaRu6fjLAl7I5DhdZEm3aEvA3Eu63jRNnPbd2djMhsaHQaM7KNXsYpmYQw1vUga78M5KX3-pJvlG0JnSuHrUN4KxyMJ3sMJpYJxbrqtpewLnL2JuejIB0hvYw=)
18. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEi-Tl4LnHXWiyh58-1X1LJxDK5DThiPgaAFhkI0xToVGCG1fJ2WsnD3FZdamrwddS2kro_Mbzn6VCG_PbbkrLwdfEAwH1ACEfbRPy4WdKdcy1CezhAZx4xn5nZUonngkHNBATHvmrf)
19. [frontiersin.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGkCvuQvJ7InCXLBPMqnEp0kWTC4u12dfmQx2bdEoPJhr3H0vaKSB3D8OI8l5uP0QdsDx7mPqpkswGpkzQb2sFVakBtOfcgqFCf0cByF_1_F85NrR5fNClUT8P9qHGtIT_2a5w_duNMfl_hEJB64pkwljWNgjjGYw2S1syJPrDEzHOKtLicbkY8jcG1o3o=)
20. [frontiersin.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGlriEgivYGVXFJeRu-10l7ewoMFyzeVlKoJ68st55V5uA7C0mpUvHAtVpxo9ekXfkTKf22haZbVoYlcjKyOfHVY9FYeKtPenv0yTM9mut54iCyOminFwgOruylrLCEW9l-PjgGgj52KJI58Cx5UrwdfxWaPBck2Mi0_V1Qrqy1nm4RtLs2wACwQwRlrpk=)
21. [jci.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHshA0wfAtjJFpjkHrU0yt0nhWJfM24tmWzuabJ0vZ-flSCg8lF5OcG9IHw_DVKxaifZ5byvFXdbHbkBAcOXH1r_u76FgJK-OsghsklYIdtE7azLUv6wh658pi8Dj7ng_Gya0hjQNkifyCX)
22. [biorxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQH5igmf-vksYLAMLYrrLYEmcUKpAkKXtzK7QJYfvj136jBXQ0kVo4gq4kmwek0fty3hq3V6dVnGp6yd5UOEN4698A-4CDFe1tdLY5jJdM9drFvvGUb1DUmqzVXYq0EiZOOkc2SvXlUZZ48EfWT_OZwu)
23. [frontiersin.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQF3RZ-EVIQpn4KOE3yvFXQJKOdV4dJNkb4kqrZCcMkj9swoYM7-rY6KhfQvnYXhkkCBXaCvIp04YE3AHlITCIIKAaWOLT3rlm7mUAMfkRZYR6xvfxKtehJrkAuKOJUGKA7ZH1ezmQLkJZM8bqIgPUq6Y-UHqweghYD8pDxmWoQbIeYClbcc9Od0EUn2IA==)
24. [researchgate.net](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHgZPodd_7D0zCgKW3Qxj8Yi9XZpnbv8iJZl0pTc7Htuo0lm-Wf1qd_J8zYxDybTTW6-vE2Lhq9ZvEMbEQ4BiLcln9Dax7VyWq0ne-etXEKVVKJ4NOHQqAQKNTQh18sY3qX4zilabqVxrLdmF0il3Avz54lJ0_AlBT86_XIOipYpwAA8izt8A3ctpRYxzJzeSwsIl0ldFQFUGLWs5EegVxqcg_MSZCN4dqcsZSIxknXBJaBzGNVzs55_yXI)
25. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEvd3WWyVpHqOb_OMPtAGF8gNZQ4tJUsk7UE5OSMQ2BOAGA3wMEyaKFY2LN3JFfa5WbpFBzW3dAyJQME1dAfoG2OTl912y6lNo7W2UpfD92Aj1cl66GRW9ML-DDzTjwsxBNutbsS-A=)
26. [elifesciences.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHqgnpJR8M7uZ60QoxUhH5QbrwFn4MQySPT6YF4dUEWq-PBJEcGpevA1W_J1MhSpsuoDeua_znkCmhw5xcp2cqMY-Bd4bLmcUYFuEXrpz7_JfWrDPMiLBF-1gSaNPfk)
27. [Link](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEzomWryS6SsL3uzidYkGVko357cvHtgTkRJsomiDqYgEulyg7gRJF2GPYVq81QTfbcfojeSVJU6bX9iGh94kKFbksLkuNKvrunFLY-1KSr8oeVHsCeEGEUsrFnhLSyq1SsvbsIjkTC)
28. [Link](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEA0CYSCr4wUpkayEqkH4-JajM_UyuV4gssUwZZzusn1A4KImXID9eQhX0l4FqRqDB7k8BDXxyvMz7_sukOndbUnhf6bEtk5MRLemE4Z6H95uFNqgBLQJawPmpiU6SU)
29. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEdQqttWL6L_4EPfFb0AprYe1gIyVQoPlS_n5MhyxjdzjJ2QcksPCbcjCUmFrVdFdB2vv6JcN2fAQBOp_v4mBJBOnCw8JH6_uCDGoTpT_1_1utzEKEV0Q09IORxLHcchvOGokp2qVr5)
30. [physiology.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGe-e1Ykr1CyaNcUfFk6fWTwFSNOCLYuFg9oiJQFwJlJSfLhe9Zlj1qFTJpuBaKgY9yyMe4xkCRREOokewu0XJzYeSCDKK1oX3VpuGwmVSBxm0WoNhEj1KUUJpSJTfRqKWQaFrwhmChQbDc_vnLpYq3qEhN70J3USk=)
31. [elifesciences.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEc67mUBIukMUhGT2NPGFvP23Z68NhWxMjlI8SwjgBpcXhdsF4f8sfYzQzL5ZXW6Md7BC1pHRSH9pnaWcnEyNlSLNVOI3rbNBEw1pwh0HfIj-_ki43THkrQ6N2tsfmt)
32. [oup.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHq-F3lVuTt5rt8yfiotryoSHN8ecifnxtc7YkQknBZkaHW03YcBRCsaLUipZ7gNpO4eCAejjFxgYf9fkBQNt8qKcCNKf_mjQaVj6LojaDVxpYxxnauxIIZNDsR--rAtzPgWyRReLLmZONZ-oO3EaP_zotGwXk2sLJ9P15du15C2g7xHkT2Yj0dfg==)
33. [frontiersin.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHGSFncyKk2haf-9VRIbGySQl9Gj-Qp4ekdJY_jSAiGDKTcNBxze6922l_4yrwgCFo_R4KsyHTsDktyuBQM5Hiw1GdnmMEbcU7A7coQV73-EdRcHauoJwaS9o173znySoO6Cn640wDd92PDPvzqc7tZ4i6kOsqLZi_Ed7CgWPZwl0rETZhM6OAdoyTQdQ==)
34. [esser-kahnlab.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFLnei2f7dJclUfAzq__oRLpzlf3nrIKTbZty-ICvw-oCcN7E2et2gek3WEWQejiIAv2X1ljTMwrxAiE9xogpHCHEkZLipy4SbapM0zH0Gd_SE_Q0j90y8x6qXd2CE_czH7XHM=)
35. [Link](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEZrPOv8NEYakQwkgBzu0MQXb5CPSpgLul8YPwGn4__nqK_ZgfWlxPqKhUZYhKKb6kVDo6-v2KsLdKG1sbJaAJe6c6zjuFjKCy6y9TN4EwXJ-HBVB4CM8uaEGVxSKHT)
36. [frontiersin.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGcRq0ruuG-ZcRUo0TuuVbjCPsNwnswv6VZmHiiKE_npfApoXL7KHcCMXTV7ni1EWvIiKp4oLWVhcQZdTaZO0n_mytrSTCl97x18DTaiHUpVUWA9F0Isy1rT1B1jD3gD4FkE-Y8pZW7yQ52cttoEd22oEl2pYU0VzYFRIMuZvwjfxRILBgfv3zBahz9Svs=)
37. [oup.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQF7bpLdNp31IXm-xpD75wNP4N6v473GC27uB8TyazHGWPIwoRL22NWpAEq8l5dCXtHosk76KgLbnv0bRpWasCALhwiJhzPXrknyGwkr_KFplBzh3O_EHCgBcgNB64Wtj1bkTziY1kvlClZCjIy9TJxK8_vK74Gn1oqMK5ZC-fkK)
38. [wjgnet.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFYso4OYpi_xuocNmfeAvLpxbhRAkwP5U1Zk_f1GOeoYNEdTpkg3i3pvGqW4nlKFXrd1LFAxE_mUUX3po4yT-McjhqSMgXQ2sdgUXZASvqDL3ziquykjmw7nqRMjKP73dEpAvil795lcaWREvI=)
39. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQF1_sl7lwZwJaPA4jnetFyUw04uGqwOiFWhZv1OKP0-u8CEmrt6oDh82LbrAq2ifC_J1X17qbVaajT0WigIEjZQ4qSsyzlsJ5johXKiK06FByCz3yh2QjODOrnpsyz3GSmvY7yq98dU)
40. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFHhJVYNJmQ0TsY16ed4bY20cERGc8rWXsztbgHVIWjPhc_NlroBuk-YxKp6sIQkQVb2DObinAEYMz12zvFoXmZ_S9Iw7inYmIwiKZO7oY6HeHVzhEmUirJv-NqsNQmNdqwABbW2dPV)
41. [researchgate.net](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGfE_k7Slu2vdNIHNpbrvQviMPhVm8pmku7RLvhcRCoUNMJq1CdPqVT4o56_sUOeLEJhXQlwI5dtfoivknAPurf4p_jgoU25k0k2_ccdK1YZVlqzILHgOHkPiuCLkbfVXouSQM_zuiZH33Zo9HrI9Te8OrsDzok4yS20HUCd0UfQENmjjv88Bo5OZ1xXLwyRpp0MLUUHtJYJ1PPbiEvXzU3Vg==)
42. [biorxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGUnPOSofDtNDnUszR4r4snkc3UKilpuiO8v1lsAjrK1IbtTZTY_JOFvmDBP7S88dTLYghU9dDRa6Lvd1djFcig3ytPwYP4y9anEgLDlxRtuwuStsiu-9UiRpeDL5LJk-dHiWIgRJXxo5SexMWB8VJzZIiQt5qo)
43. [pnas.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEgiotSesGcbusa-rCt3xfldAWMmrtAQKdyruiK3pvA4waVihK_V2-nUeLx28m2jQssmWl3mQVtlUqVHRitoGFY0lZQRX9nYtf069Msrh6owb0XcCmoDPxcUA33WkttgAtb_UGYgg==)
44. [biorxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQENREb3MdgLfaSqlpOnLhuiBvJRmnhDoZ-cecxseSGgk43cS94SJ5Or2bo3AVnAWCavIEAqEIAmDlRZikVKr9T8QuhfLDsGHoEaU8lqqBDoDoV_Ijqk1KvHV-FKChb-uXd3KI-0ZOQOGq1qD-DVMo9OYPJAdxbqCflx-g==)
45. [mdpi.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGNC_nYaMq-QvgTRJTEWMrgt_GZW3_V2ec8vXIvEOSRDKGqz3nSK_LXrcy52cFd0uFyzYcG2EJZx6YKLbqqGnCW_UwwtfjVJriBwipQpzqP_xlRnUiwrpvqrD35sg==)
46. [jci.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHMjv-YBRn3dwb6HEo9ck-MhZiJILEXK6dXdQ8jh7nj8obAAHy4_9sI9bGTPlnOjttrLxaDB0Z032qtYbOEp-889bTZ9xx9EOmayj_8bFVPlhx8sKfhy7m72i9-xbU=)
47. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFmOd-99RC7oBfyEKppZWR52_L0M5Hm3nLsv8Sp75bYNs-UQaAPHbnR1CYqUMXF8gmbiksur6m4byIaSzxG-m4OBqN8tECXMxDT-9et_VnYAxsme6B6xGz3hT1cZ152NUunjNOyWwMx)
48. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFWibdkbhkcVeR1V5MSbb9AgoBqlLSKeT7foHxVE5h6xVXknNOIVCvK6hCnpYH1fyS14cPYXH0nTslH-QdiYz0E3554dhrLYdxdpDouHOdBtMig0xDLXTmjCgUzgI4rVXS1B1P9wO5h)
49. [frontiersin.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHSkor9wMyLkLnVuRBP4i5APXBuT_XVOml9ER6ncQYHO6jaLOr8ttdPVed4k2qTUj2zJlsJAVPJKvUVdpt96evIRYKbjH7c-KWlC6HUHzIiZJ7iVOHYKDrm5KmRmI55r01X11bOhUVZxtcm9mQ0j2fofLLTipvCJDsVYlu0BRqlHyCOieAftuqPrTignro=)
50. [oup.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEPF0uWyfpLVAQiqc04AXfk880XhmukEyRyVy_uKXnrv6kZywDxHVBZWMViz9AAvPJUcyF7vnaz1lb5Xr5YIbqbF0spLV2rhQ8sV7EFw7_Y78WYukvpnMBzo9HX0RT5E--8ugrIqlYyWBvu5TJv)
51. [pnas.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHI4JojU-1Y38rN2tFUeQhBVIFr3eYr8zDXVmV-EM9III3lrntSxjqcmBf2eGIMyPAKbQZpLFA5UMpqOdGMdUzwaufIXqEbQui2Ft5xIVI3yyfEplQNOXdTfTvQ7Tq1IGQs_GgkwA==)
52. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFQgJJVpjSVeBiqnI4MPcod6kXRN0cOuInWa_6qgQZebVPbIRdeeI_FqmG0MzMbK2IFGVP-m2CptgiX_j02-BaUQznktvFAV8yeSCdotUPTvXWWJ5YqeG3IftIyCOnZ3iCAhj-FbaY=)
53. [mdpi.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQG6dAM-MgrM9lWaIe2viO-0AJnnSB0mm-JrMDY-m__UYiGpHZAD_JU8EN4FZKHyopM0kBCE2VES_9850zjr9qLROX5AhqUE_jYgxSMB2grtNCVIpMnm9scGS66ubA==)
54. [scienceopen.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFyXW437c9PLFG9J_3CUompEWQvpNT-6wN_2pk4AyEzPfTmijXDPdFxvBZ1sXDmC3m7P6Ter9E4xbNUxOQn7XB7lVSVkHfTKDHp_FhL7ehBU0thjdYgIRgBJXTLw8k4tHtH_R7ZZR7zVT8qgucvjgKWzSG49Opyzzd7By0Hew-YDbh7aHMuGG6zoXs7SB0xGpBZRDAt8zmzk0ck5oo9evmyu25dWhVNTHg6nEwlKwzpoU0OvCHewOKt2ZWD)
55. [news-medical.net](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHZK4wxY4fFd3eF-vza3ytKs2h7MTn3Z-adZpleuLcSCGYuLSfu_RVX6uFPwkIUlGCfS3Xm3Nj1gGgEaavUkyyiTZOb2wb3Ut0ZNxEHuSmh4RM4bMgrRQtw0g7JgyzkVxhvpEok7AI3lWAJ10edba1mcAKV4lvWDryvQW9kD_P07lepNknf9IAxWi4aNkkvmoIWsXxiXAarLkc7ta6QRHus1oD2CE3d)
56. [grc.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHvJbcWyFFuPd4tkR7lNOMnwd_Xw0QbXDP9Sp2ef2R26JUDQEhGLPACKK3HlodxPYupI36209bYKIJkX_lIPllwxFCFdhhtm36B7wjKHs5ioGXYW87_afBb3vuW9ZBpJDu-5hgscKyxPp1UvBFCquO5cuFFlrpmBa4XCEzYObH0kEBDpxReG0Cuqg==)
57. [frontiersin.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGk_gxuk6A2h-YQSJma4an94YuwOKcj1h4_TUKoseMnp2yfqHcFkFf3SK6jkXhA7BL0vKGfup7IKTG7GqkBLajY7eqAGYVSq-GCb6BNwnjdSakmdHV1SLtAoUE_xlXAR3syqLuIpekr39HNCCJov92RyVyh9jSnhYXOXYCN77hApkJBqHobySLchYLY)
58. [news-medical.net](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEJIE8BCxZtI0Y124FYv7O7B85D0urKUsffELVCuSReksfwylygNwvd16R8lQ9-L9ONEog1vvEMzLisPkwsBy7v-5v0b3oNxTpxAGMgCdxLpwk7Zb0WY9toLmv4cfMIXYjmnED-9OObY799AQM4p4f_c-mpcxMBPbVqo6Lkw81A1As2YGp0FI9ibyUgxIXqJ1nVu9s_CC8hsIE_Z-TF7B-QFZW_reGjR6TBi4IrjCh6PTpI)
59. [ahajournals.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHXvzSoUgLz0vdYzvOxmgtDUoIy_mBY2rtgyPLK6TJEhATKOrtlnjPLmytxhjBErxXLArbdgPF4oootp0dfijsF96sNq031pDPkF_XC_wf0RO39-7oyUenBqV9ZJegCg25-j1mQwUdSD14LljKg2RoDVEks5ENW3Q==)
60. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFoN92UbqCqhXXKkGawMC5uLRYeFy2DW3CeWZr1CywEsAQV0k3QrQMm9dMBP8bkTRBPnVPOg2SIkn4Y5O7_iP-0Q6r3PN8qEQaeTEIivT-_TvCR4yBYhW9gR9xYV5IhvD2N4X9K9xbi)
61. [wikipedia.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHz9H5aN1E-7zs4JqGN_TLJgtwc7fH5t6GBnVAyKNXlnJTwm-0sAK3go62uFrOvSm4Qz0YhA1-omvOfgO9Zhf73emmBiXuBdE5pj6e50XHHq1oxqnN5vSu40n9wIovdajVZ)
62. [wikipedia.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFbpEajG9Jjx-fV8RU9XR0COdFPGLVj0dFPWsGtkhfGyVnFdlu8-FrQz9Gcg9knOUj8zGGOYkVhvKL_zFgYfgoCHWNswKj4RF2Pz3XKWg2B_Pe-eLQgnelLrD_elau4ZJw=)
63. [emergentmind.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEzhdFoYjPtl0F6KXdtQ5ww-CtuKGw3lDbLokZbTn4__yBNvlO9nJnarr-61ekCnyWMOMRbSqBEfpifW6P-BGR94DoEH2oYRDpTeWMRr_JAxZtfZbz_dQkbQXUewQfLzXzFU00WhI9anJE=)
64. [arxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEOZGUK1D_KmSm0QGJWv3-DTj_7JWC_k4zv0xbwHFIDlA_5LCwCUzpMhvYl1Iuxs-cFIj4JP5QvHq7uTxl74kWaD4spfcG2_9d9sBdouSGtYlTdL9L_)
65. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHysPX5-fwFrFX6Lqgoc4PbgpMjL-iVq3YmRx8Ory9nUVqLuWRhcikE49zT2RFIMcgMjlJksKgXZJGxM3Pz-FemjU5F0iu6zxQpasp_NfP3IrfdBlOw6py0tgnbzNZxwdfTqPYBLoU=)
66. [pnas.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQH_y_j6rS6Fdlqy-IXx8yYR8a1XsEAW081t9h283fk9agvIf5tATIi7Zo5B5Wyo94pUR_xKMkKxAgDvbaM7XvA2G4MoxV2FffkK1bLABoos2-650XEmdvVItC-kL7n1Gf1eHpDlvA==)
67. [medium.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQH40IAi1ovtc0KSI79x6iqqNoVt395BkUxRreUJgn-x6HiY49H8StE51WezPTKDrnJ_Bca1upRixTt8vEIUkWcoMtDMqW4-HugALp_S_-ymRjVP1YNwTYwP3HJ29pB_VYK6QSSVu-4lqndYTLf31O58mjW1BP-3vE-5_Ssyi6x5GmalZtWaUg6ZuPqr7aceWwkDTg83dUddw7jca8wLRy-uSE_Fv_wOQw==)
68. [scienceopen.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHBM7dRHjNjun4jVyh9jXy6Og1R0hdVYOCZHpLHXJKqqzr8ntbYhkwiQfJZhM_0jYThsB0NfJCfbymMM7XlE4FX4oGKJUXxSnXfqZsjMT7GB35--L9DFyyhcDU6_idlAWJ5Cy4e7HBtglHCzM35M66a5Tn6BeglpeCpCfTQ)
69. [thedecisionlab.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEumUtrmZ3Sgsm41WfbP2FJ_q8Tgsi1XR19vabnc1gm9uX8UPq4-1ftB2eINHJt6JYVGZQS4YdKa_b15KYuiPrC7ED1Bu2jhPgn22O9akXwZbL8I08xqPkw-z4LiXaWLlslWfrtZJBZN4UOzJ6AbzBSHE_sJ7gjdcuz3oJZzQ==)
70. [annualreviews.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHR1hztaDFTRb_67-ASs-O5tMdGZbhb05-uhNSNqqTvhBVcCM5aCTljQ8xfWJDQ9VYcRz24FeTYMVcHVI4wq0KEn_bltsKiD8i63lZHHMXFPW2numpexIxyN932iuXAOHxLnagjrGkRhmtJOvJq8Ymw8hQeCrzS5Lhnki5Y4veXUKv_Et-8vS37WYJxqfrcJcHtFT3nB-JBmhSLwrBTcxsBQHze2T2rCkXrkZsRvhq6cw==)
71. [iit.edu](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGtvWt7WFqxadGGyz29T5HvSff5T0XS9fTa_Ui_uPcZr1os8bF2QM5r0Yun_wdBTwbyPyS6Qa4DJccBJ4l3FP0v9k_S7U81Lvd6cZ1rkBigJPKvKfPpmeiDE5p8DfaNIcBlDA6MuA_0A_D0Hmlsj7-qlmA=)
72. [biorxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFzGPF2EysBE5BMbUoyL6VZhjssEo4uNq3HkKwWQ1w0nDd-PYpq4D-23ICp05Q07wlia_GVCQHtuCh7fqsJuWtHCkas9rxSiii-gU_fdvyM9x4kg1vFK-TtePR823vTyxd2_h_nIEJoCjskn96aAdi-)
73. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQH3BBkf2rct--Z7Me6hx4DqvxgeS-iDk_6nQw70BuxbMbqpGXrtToeWFUQmagC2QAtXBvasC7prBwh_HKotA-VsyI3zB155O2IiXzJ7EHFWKDquJjXJmfzwl-JxOIDRrDr3n2QPJQk0)