Application of Single-cell Deep Learning in Elucidating the Mapping Relationship Between Visceral and Body Surface Inflammatory Patterns

arXiv:2503.17418 · q-bio.GN · Submitted 2025-03-21 · Read on arXiv

Listen

Transcript

Introduction to the show: ident: Genomics Radio. Generated commentary on the latest computational biology and genomics papers.

Ines: I'm Ines.

Marcus: And I'm Marcus, here with Yuki.

Ines: We're looking at the paper: "Application of Single-cell Deep Learning in Elucidating the Mapping Relationship Between Visceral and Body Surface Inflammatory Patterns".

Marcus: The gist The study reveals that single-cell deep learning applied to subcutaneous fascia (scFascia) can elucidate the mapping relationship between visceral and body surface inflammatory patterns by identifying a fascia-visceral…

Ines: First, who's behind it and why it matters.

Title and authors: Ines: We're diving into this paper titled "Application of Single-cell Deep Learning in Elucidating the Mapping Relationship Between Visceral and Body Surface Inflammatory Patterns." We’re going to talk about how they used that deep learning approach to map inflammation between the gut and the body surface.

Marcus: Yeah, these authors are Haixiang Huang, Bingbing Shen, Zhenwei Zhang, Jianming Yue, Lu Mei, and Qiusheng Chen. They focused on linking visceral inflammation with what's happening in the subcutaneous fascia.

Yuki: I’m curious if this pattern they found in rats really reflects what we see in larger human populations regarding systemic responses to enteric infections.

Ines: Well, the core finding is that visceral inflammation can actually change cell states located far away from the source of the inflammation. They looked at how *Salmonella* dysentery caused a shift in expression profiles within the scFascia cell types #pg3.

Ines: Essentially, they’re showing that instead of just a local infection response, there's a systemic communication network connecting the gut and the skin tissue through this FVICP.

Yuki: So they are mapping out how these two different environments are communicating in response to an infection rather than just observing them separately.

Ines: Exactly. The study shows that macrophages in the scFascia exhibit Warburg-like metabolic reprogramming when they're dealing with dysentery, which means their metabolism shifts toward a more proliferative style of energy generation #pg3.

Marcus: They also used PySCENIC analysis to infer transcription factor activity from the single-cell expression matrix and found that three transcription factors, Fosl1, Nfkb2, and Atf4, are significantly upregulated downstream of Wnt/Fgf signaling #pg1.

The paper's summary: Ines: To summarize this paper on "Application of Single-cell Deep Learning in Elucidating the Mapping Relationship Between Visceral and Body Surface Inflammatory Patterns," they are showing how visceral inflammation can signal systemic changes through remote tissue cell states.

Marcus: Right. They established that the scFascia expression profile is a sensitive indicator of systemic homeostasis, which is really important for diagnosing things where you don't just see what's happening at the site of infection.

Ines: The key finding they pulled out is that this isn't just some random signal spreading around; it’s a systemic spread of signals from the visceral organs interacting with the whole circulatory system to affect remote tissues.

Marcus: And they showed that the immune state in certain visceral cells mirrors what happens in those distal scFascia cells, specifically seeing this Warburg-like metabolic shift in macrophages during dysentery #pg3.

Yuki: That metabolic change, where they downregulate the TCA cycle and upregulate gluconeogenesis, it really suggests these macrophages are adapting their energy use because of the visceral stress.

Ines: And what makes it interesting is how they used deep learning to pinpoint the mechanisms driving this; they found that CPTCs are actually sending those signals to the macrophages.

Marcus: It’s not just that CPTCs are there, but they're showing robust signaling activity toward them through things like Wnt and Fgf pathways.

Ines: So when you look at the data, they found specific ligand-receptor interactions—like Wnt2 over Wnt5a and Fgf10 over Fgfr1—that are key to this crosstalk pattern.

Marcus: And then they used PySCENIC analysis on that single-cell expression matrix to infer which transcription factors are getting turned on in response to these signals #pg1.

Yuki: They found three specific transcription factors, Fosl1, Nfkb2, and Atf4, that are significantly upregulated downstream of the Wnt and Fgf signaling.

Ines: Those targets they found were heavily enriched for functions related to cellular proliferation and DNA replication in the tissue #pg1.

Marcus: So when you put it together—the CPTC signal activating those TFs—it directly explains why you saw that augmented proliferative potential in the macrophages.

Ines: It really confirms their hypothesis about CPTC-induced signaling driving the metabolic reprogramming we saw in those single cells.

Marcus: And for someone listening who doesn't work on this, it means inflammation isn't isolated anymore; it has a systemic reach that can be detected by looking at the tissue outside the main site of infection.

Ines: That’s right. The paper proposes that scFascia acts as a sensitive indicator of systemic homeostasis under various pathological conditions because its expression profile changes in response to visceral signals #pg2.

Marcus: And they even showed some predictive power; their models trained on each group performed very well when tested against the opposing group's data, with one model hitting ninety-six point four four percent accuracy on a test set #pg2.

Marcus: We’re still looking at how this specific pattern—the FVICP—changes depending on the type of pathogen or the specific tissue involved.

The paper's improvements: Ines: The authors suggest they can take this work and make it even better by focusing on improving the AI system for disease diagnosis, suggesting it should be trained on whole-cell expression profiles from scFascia samples to predict visceral inflammatory states.

Marcus: That sounds like a big step up, moving from just correlation to actually predicting a state based on the tissue itself.

Ines: Exactly. The paper shows the current models generalize well, but they propose an improved AI that can act as a sensitive indicator of systemic homeostasis under various pathological conditions.

Marcus: What I'm hearing is they want to build something where the scFascia profile becomes a reliable diagnostic tool for visceral diseases because it’s showing this sensitivity across different groups.

Ines: And they're also pushing the idea that this improved system can map the relationship between visceral and body surface inflammatory patterns by specifically identifying those regulatory mechanisms we talked about, like the FVICP.

Marcus: That means when you look at an scFascia sample, it wouldn't just be a data point; it would be part of a network showing how visceral inflammation is spreading outwards.

Ines: And they want to specifically track those CPTCs again—how they drive macrophage signaling and the subsequent metabolic reprogramming—to understand the mechanism more deeply.

Marcus: It’s about integrating the transcriptional data with that functional insight, so you get a picture of not just what's there, but what’s actually driving it.

Ines: The implication is that we move closer to having a tool where we can see how systemic disruptions affect localized tissue responses in other disease models down the line.

Marcus: It shifts the focus from just finding correlations to building predictive models that can actually tell us what’s going on systemically before it becomes severe.

Ines: And they are essentially saying that by using this single-cell deep learning approach, we can move past just seeing a reaction and start mapping the underlying communication network.

Marcus: So, the next thing we need to watch is how robust these models are when we move them outside of the rat model and into larger human cohort data.

Conclusion: Ines: So to wrap up this paper on "Application of Single-cell Deep Learning in Elucidating the Mapping Relationship Between Visceral and Body Surface Inflammatory Patterns," it shows how visceral inflammation can signal systemic changes through remote tissue cell states.

Marcus: Right. It proves that the scFascia expression profile is a sensitive indicator of systemic homeostasis, which is really important for diagnosing things where you don't just see what's happening at the site of infection.

Yuki: From a population perspective, this suggests that we might be seeing a more unified systemic response to enteric infections across different hosts, which is what we need to consider when looking at broader species responses.

Ines: It changes how we think about inflammation because it’s not localized; it’s a mapped relationship between two very different environments.

Marcus: So the next step for us is figuring out how to translate this into a functional diagnostic tool that can actually be used in a clinical setting, not just in the lab.

Yuki: I wonder if we need to look at these same mapping principles when we study inflammatory responses in other complex systems beyond just gut infections.

Ines: Exactly. We’ve got a lot of ground here on how single-cell data can reveal hidden crosstalk patterns, but what’s the next paper going to focus on is actually building those predictive models for those specific states.

College of Veterinary Medicine, Nanjing Agricultural University

q-bio.GN

Submitted: 2025-03-21

Updated: 2026-10-19

Comments: Published in Computational and Structural Biotechnology Journal 27 (2025) 4183-4200

Journal ref: Comput. Struct. Biotechnol. J. 27 (2025) 4183-4200

DOI: 10.1016/j.csbj.2025.09.040

Code: https://github.com/aertslab/pySCENIC

License: http://creativecommons.org/licenses/by-nc-nd/4.0/

Importance score: 83/100

The gist: The gist The study reveals that single-cell deep learning applied to subcutaneous fascia (scFascia) can elucidate the mapping relationship between visceral and body surface inflammatory patterns by

Key concepts

Fascia-Visceral Inflammatory Crosstalk Pattern (FVICP): This pattern describes how inflammatory signals from internal organs (viscera) spread to and affect distant tissues like the subcutaneous fascia. The study found that visceral inflammation triggers a corresponding pathological equilibrium in the remote tissue, linking systemic inflammation to local changes in cell behavior.
Metabolic Reprogramming: This refers to a significant change in how cells generate energy and process nutrients. In this context, macrophages within the scFascia changed their metabolism during dysentery, shifting towards a Warburg phenotype (increased reliance on glycolysis) and altering their TCA cycle activity.
Deep Learning for Biomarker Discovery: This involves using advanced machine learning models to analyze complex single-cell data from scFascia. The AI successfully learned patterns that allow it to predict disease states with high accuracy, proving that the transcriptional profile of this tissue can serve as a reliable indicator of systemic inflammation.

Summary

The gist The study reveals that single-cell deep learning applied to subcutaneous fascia (scFascia) can elucidate the mapping relationship between visceral and body surface inflammatory patterns by identifying a fascia-visceral inflammatory crosstalk pattern (FVICP) in response to Salmonella dysentery.

Systemic Impact of Visceral Inflammation

Visceral inflammation can influence cell states located distally on a significant scale, as shown by the transition of expression profile in scFascia cell types following Salmonella dysentery induction In this study, Salmonella dysentery induced transition of expression profile in the scFascia cell types, proving that visceral inflammation may influence the cell state located distally on a significant scale. Each cell type in the scFascia served a different role in the view from the deep-learning framework

Identification of Pathological Crosstalk Patterns

The study established a pathological macrophage-associated microenvironmental homeostasis, akin to findings in prior studies, where macrophages and CPTCs are frequently localized in the perivascular region forming a composite structure alongside pericytes and other fibroblasts Deep learning reveals impaired aerobic respiration in scFascia MPs and identified potential crosstalk with CPTCs in dysentery. The intercellular communication network within scFascia under dysentery conditions exhibited an upregulation of macrophage proliferation and metabolic reprogramming mediated by CPTC-derived signaling Prominent ligand-receptor interactions included Wnt2/Wnt5a-Lrp6, Fgf10-Fgfr1, and Rspo3-Znrf3

Metabolic Reprogramming in Macrophages

Macrophages (MPs) of scFascia in the DG exhibit augmented proliferative potential and Warburg metabolic reprogramming This MPs’ metabolic alteration mirrors the Warburg phenotype observed during enteric bacterial infections, typically indicative of adaptive homeostatic adjustments in response to enhanced cellular proliferation The analysis showed that MPs exhibited significant downregulation of the tricarboxylic acid (TCA) cycle and upregulation of gluconeogenesis pathways under dysentery conditions compared to the healthy control group Furthermore, a marked reduction in the proportion of cells in the G1 phase and a concurrent increase in the S phase was observed under dysentery conditions compared to the healthy group

Regulatory Mechanisms Uncovered by Deep Learning

PySCENIC analysis inferred transcription factor (TF) activity from the scFascia single-cell expression matrix to explore underlying regulatory drivers Analysis of MP TF activity revealed significant upregulation of three TFs downstream of Wnt/Fgf signaling in the dysentery group: Fosl1, Nfkb2, and Atf4 GSEA functional enrichment demonstrated that these target genes were significantly enriched for functions related to cellular proliferation and DNA replication in the DG, while functions associated with aerobic respiration were markedly downregulated This result validated the hypothesis of CPTC signal-induced upregulation of MP proliferation in scFascia of DG

Diagnostic Potential and Model Robustness

The transcriptional profile of scFascia cells exhibits significant potential as a diagnostic tool for visceral diseases Models trained on each group exhibited exceptional generalization performance when applied to the opposing group’s dataset, with the G1 model achieving an accuracy of 96.44% on the G2 test dataset These results underscore the robustness and reproducibility of scFascia expression profiles as a reliable biomarker for disease states The study proposes that scFascia, as a pervasive connective tissue, mediates the correlation between superficial tissues and visceral organs serving as a sensitive indicator of systemic homeostasis under various pathological conditions

Key Findings on FVICP Characteristics

The FVICP can be concluded with the following basic characteristics for the first time: 1. Systemic dissemination of visceral pathological signals: FVICP is underpinned by the systemic spread of pathological signals from visceral organs and the extensive interaction between the circulatory system and the microenvironment in remote responding tissues 2. Synchronization and mapping of immune states: The immune status of certain visceral organ cells is mirrored in distal tissues, such as observing Warburg-like metabolic reprogramming in scFascia macrophages during intestinal dysbiosis 3. Pathological homeostasis in distal microenvironments: Remote local immune microenvironments establish a corresponding pathological equilibrium in response to visceral dysregulation

Methodological Details

The expression matrix was generated using Cell Ranger version 7.2.0, and the primary processing was executed using Seurat v4.4 within the R environment version 4.3.0 The optimal model parameters were identified through a hyper-parameter grid search strategy, ensuring the identification of optimal configurations for the deep learning framework Cell communication analysis was performed by applying quality-controlled Seurat objects of scFascia and colon from each group to generate Cellcall objects following the guide of this GitHub project The cell cycle distribution within MPs demonstrated a marked reduction in the proportion of cells in the G1 phase and a concurrent increase in the S phase under dysentery conditions compared to the healthy group The study utilized Transmission Electron Microscopy to investigate potential functional interactions between macrophages and CPTCs, revealing that morphological evidence of functional interactions between macrophages and CPTCs was observed The researchers also employed DeepSHAP to extract interpretable features, including key signature genes, to elucidate the underlying transcriptional dynamics The study utilized PySCENIC to infer transcription factor (TF) activity from the scFascia single-cell expression matrix The study also employed an MLP model for binary classification tasks and a simple yet highly accurate binary classification task using an MLP model

Experimental Model

Six female Sprague-Dawley rats were subjected to a 3-day oral gavage regimen to induce Salmonella Typhimurium infection The bacterial suspension was adjusted to a concentration of 5 × 10 8 colonyforming units per milliliter (CFU/mL) using 0.01 M PBS To neutralize gastric acidity prior to infection, the dysentery group received an oral dose of 5% sodium bicarbonate (NaHCO3) Specimens comprising intact skin and scFascia along the abdominal midline with a width of 2cm from the pubic symphysis to the xiphoid were harvested for paraffin sections and transmission electron microscopy analysis The scFascia was sampled along the abdominal midline with a width of 2cm from the pubic symphysis to the xiphoid, meticulously dissected to isolate it from surrounding dermal and adipose tissues Dissociated cells were collected at 20-minute intervals to optimize cell yield and viability The resultant cell suspensions underwent sequential filtration through nylon cell strainers decreasing from pore sizes of 100 µm to 40 µm to remove debris The remaining red blood cells were lysed using 1X Red Blood Cell Lysis Solution Cell viability was assessed using 0.4% Trypan blue staining and confirmed with a Countess® II Automated Cell Counter The study utilized a water-in-oil solution to prepare suspension cells for 10X library construction The expression matrix was generated by Cellranger 7.2.0, and the primary processing was executed using Seurat v4.4 within the R environment version 4.3.

Improvements for AI systems

For disease diagnosis, an improved AI system can be trained to predict visceral inflammatory states by analyzing whole-cell expression profiles from scFascia samples, as demonstrated by the models achieving exceptional generalization performance when applied to the opposing group’s dataset. This system can provide a sensitive indicator of systemic homeostasis under various pathological conditions.
The improved AI system can map the relationship between visceral and body surface inflammatory patterns by identifying novel regulatory mechanisms, such as the fascia-visceral inflammatory crosstalk pattern (FVICP). It can specifically identify that CPTCs displayed robust signaling activity toward macrophages and track the subsequent downstream effects on macrophage metabolic reprogramming.
The system can perform predictive functional analysis for macrophage states by integrating single-cell metabolic data with transcriptomic features, allowing it to predict augmented proliferative potential and Warburg metabolic reprogramming in macrophages under conditions like Salmonella dysentery.

Abstract

The subcutaneous fascia (scFascia) is a widely distributed connective tissue, yet its contribution to systemic responses during visceral inflammation remains insufficiently defined. In the rat model of Salmonella-induced dysentery, scRNA-seq of scFascia combined with deep neural network (DNN)-based nonlinear feature discovery revealed transcriptional alterations indicative of Warburg-like reprogramming in macrophages (MPs), accompanied by reduced activity of the citrate cycle. CD34+/PDGFRα+ telocytes (CPTCs) regulated MP differentiation and proliferation through Wnt/Fgf signaling, suggesting a pathological communication pattern between scFascia and visceral organs, termed the fascia-visceral inflammatory crosstalk pattern (FVICP). PySCENIC analysis showed increased activity of transcription factors Fosl1, Nfkb2, and Atf4, modulated by CPTC signaling to MPs, leading to downregulation of aerobic respiration and upregulation of the cell cycle, DNA replication, and transcription. Co-culture assays demonstrated that CPTCs suppressed the pro-inflammatory activity of MPs by transcriptionally downregulating TLR4 and CPT1A while upregulating ATF4, and enhanced anti-inflammatory potential via post-transcriptional MTOR upregulation. These findings highlight the role of scFascia in immunomodulation and metabolic reprogramming during visceral inflammation, emphasizing its contribution to systemic homeostasis.