arxiv: 2604.24796 · v1 · submitted 2026-04-26 · 🧬 q-bio.OT · cs.LG

Recognition: unknown

A multi-stage soft computing framework for complex disease modelling and decision support: A liver cirrhosis case study

Xueyuan Huang , Yuheng Wang , Yuanzhi He , Siqi Gou , Lu Bai , Wenqian Wu , Peifeng Liu , Aijia Wang

show 3 more authors

Tianhui Fan Ze Zhou Jiayu Xu

Authors on Pith no claims yet

Pith reviewed 2026-05-08 04:53 UTC · model grok-4.3

classification 🧬 q-bio.OT cs.LG

keywords liver cirrhosissingle-cell transcriptomicsmachine learningconvolutional neural networkgene signaturedisease modelingmolecular dockingendothelial subpopulation

0 comments

The pith

A multi-stage framework using single-cell transcriptomics, network analysis, and CNNs identifies an endothelial subpopulation and seven signature genes in liver cirrhosis while outperforming conventional machine learning.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents a multi-stage soft computing framework to model complex diseases from high-dimensional, noisy, and limited biomedical data where standard methods often lack robustness. It combines single-cell transcriptomic profiling to locate relevant cell subpopulations, high-dimensional weighted gene co-expression network analysis to stabilize gene modules, restructuring of tabular features into two-dimensional disease maps for convolutional neural network analysis, and molecular docking for therapeutic evaluation. When tested on liver cirrhosis data, the framework pinpointed a disease-linked endothelial subpopulation along with seven signature genes and showed stronger classification performance than conventional pipelines. A reader would care because the approach is designed to be reusable across other diseases marked by data sparsity, correlations, and heterogeneity, potentially aiding more reliable discovery of cellular and molecular drivers.

Core claim

Using liver cirrhosis as a representative case, the framework identified a disease-associated endothelial subpopulation and extracted seven robust signature genes (HSPB1, GADD45A, CLDN5, ATP1B3, C1QBP, ENPP2, and PARL). The CNN-based representation learning module, applied after converting molecular features into two-dimensional disease maps, outperformed conventional machine learning pipelines in classification. The overall framework is presented as disease-agnostic and extensible to other omics-driven biomedical applications.

What carries the argument

The multi-stage pipeline that stabilizes gene modules via high-dimensional weighted gene co-expression network analysis and then restructures tabular molecular features into two-dimensional disease maps for convolutional neural network processing to capture non-linear interactions.

If this is right

The framework supports identification of cellular subpopulations and gene signatures that can inform targeted therapeutic exploration through molecular docking.
It offers a reusable structure for other complex diseases involving high-dimensional omics data with limited samples and feature correlations.
Conversion of features to two-dimensional maps combined with CNN analysis improves handling of non-linear interactions compared with conventional pipelines.
The multi-stage design yields both predictive classification gains and interpretable outputs such as signature genes for downstream decision support.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The 2D disease map conversion step could be tested on other high-dimensional data types such as proteomics to check whether the CNN advantage generalizes.
Independent clinical cohorts could be used to check whether the seven signature genes function as reliable biomarkers across populations.
Linking the identified genes and cell types more directly to longitudinal patient outcomes might strengthen the case for clinical translation.
The modular stages could be swapped or extended to incorporate additional data modalities without redesigning the entire pipeline.

Load-bearing premise

That high-dimensional weighted gene co-expression network analysis reliably stabilizes gene modules under sparsity and noise, and that converting tabular molecular features into two-dimensional disease maps enables a CNN to capture non-linear interactions better than direct tabular processing.

What would settle it

Running the CNN classification step on an independent liver cirrhosis single-cell dataset yields no accuracy gain over standard machine learning pipelines, or the same endothelial subpopulation and seven signature genes fail to appear consistently in additional patient samples.

Figures

Figures reproduced from arXiv: 2604.24796 by Aijia Wang, Jiayu Xu, Lu Bai, Peifeng Liu, Siqi Gou, Tianhui Fan, Wenqian Wu, Xueyuan Huang, Yuanzhi He, Yuheng Wang, Ze Zhou.

**Figure 1.** Figure 1: Flow chart of the study and the mechanism drawing of endothelial cells causing liver cirrhosis. X. Huang et al.: Preprint submitted to Elsevier Page 9 of 20 view at source ↗

**Figure 2.** Figure 2: Single-cell sequencing identifies key cell subpopulations. (A) Single-cell clustering results for normal and cirrhotic samples. (B) Single-cell annotation results for normal and cirrhotic samples. Different colours represent different cell subpopulations. (C, D) Bar graphs show the number of cells in each subpopulation: (C) control group; (D) liver cirrhosis group. (E) Horizontal bar graphs were used to de… view at source ↗

**Figure 3.** Figure 3: Identification of key subpopulations in endothelial cells and cellular communications. (A) Cell clustering in endothelial cells. The picture on the left shows the normal group; the disease group is shown on the right. (B) Horizontal bar chart shows subgroup differences between normal and disease groups. (C) The spherical linear plot demonstrates the amount of cellular communication between different cell c… view at source ↗

**Figure 4.** Figure 4: HdWGCNA of endothelial cells. (A) An appropriate soft-thresholding power for constructing a scale-free network. (B) Genes with similar functions are grouped into the same colour module. (C) Genes in each module were visualised and sorted by kME score. (D) Correlation between each module. (E) Expression positions of the genes in each module in the cirrhosis group. (F) Expression of seven cellular subpopulat… view at source ↗

**Figure 5.** Figure 5: Proposed temporal analysis of endothelial cells and enrichment analysis. (A) Filter conditions: mean expression ≥ 0.1, empirical dispersion ≥ 1. Visualisation of the screened genes. (B, C) Growth and developmental trajectories of endothelial cells. (D) Distributional position of seven subpopulations of endothelial cells in growth and developmental trajectories. (E) Changes in expression of LC-Endothelial s… view at source ↗

**Figure 6.** Figure 6: Machine learning and deep learning for the diagnosis of liver cirrhosis. (A) The LASSO Cox regression model shows partial likelihood deviance versus log(𝜆). (B) The lambda parameter indicates the coefficients of the extracted features; horizontal coordinates indicate the effect on the independent variable and vertical coordinates indicate the coefficients of the independent variable. (C) Comparison of seve… view at source ↗

**Figure 7.** Figure 7: Cluster analysis. (A) Differences in the expression of 25 LC-endothelial genes in the LC group and CT. (B) Correlation between seven core genes and immune cells. (C, D) NMF analysis. (E) Heat map of genetic differences in the two subtypes. (F) Differential expression of LC-Endothelial gene in two subgroups. X. Huang et al.: Preprint submitted to Elsevier Page 15 of 20 view at source ↗

**Figure 8.** Figure 8: The screening of drugs. (A–D) ENPP2 protein with: (A) Schizandrin B, (B) Fenofibrate, (C) Genistein, (D) Quercetin. (E) C1QBP protein with Folic Acid. The blue section represents the protein structure; the solid yellow section represents the binding site around the small molecule; the dashed yellow section is the hydrogen bond and the number is the strength of the hydrogen bond; the hydrogen bonds connect … view at source ↗

read the original abstract

Liver cirrhosis is a major global health problem causing millions of deaths annually, and timely detection with aggressive treatment can significantly improve patients' quality of life. Modelling complex diseases from biomedical data is computationally challenging due to high dimensionality, strong feature correlations, noise, and limited labelled samples. Conventional Machine Learning (ML) pipelines often struggle with robustness, interpretability, and generalisation under such conditions. In this study, we propose an ML-driven multi-stage decision framework for complex disease modelling and therapeutic exploration. The framework integrates single-cell transcriptomic profiling, high-dimensional network-based feature stabilisation, multi-model learning, deep representation construction, and post-hoc decision support. Specifically, single-cell sequencing data were analysed to identify key cellular subpopulations, followed by high-dimensional weighted gene co-expression network analysis (hdWGCNA) to stabilise gene modules under sparsity and noise. To enhance non-linear feature interaction modelling, tabular molecular features were restructured into two-dimensional disease maps and analysed using a CNN. Finally, molecular docking was incorporated as a decision-support module to evaluate candidate therapeutic compounds. Using liver cirrhosis as a representative case, the framework identified a disease-associated endothelial subpopulation and extracted seven robust signature genes (HSPB1, GADD45A, CLDN5, ATP1B3, C1QBP, ENPP2, and PARL). The CNN-based representation learning module outperformed conventional pipelines in classification. The framework is disease-agnostic and readily extends to other omics-driven biomedical applications involving uncertainty, heterogeneity, and limited samples.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This paper assembles a standard pipeline of single-cell analysis, hdWGCNA, 2D CNN maps, and docking for cirrhosis but supplies no metrics, ablations, or stability checks to support its gene list or performance claims.

read the letter

The main takeaway is a multi-stage framework that chains single-cell transcriptomics, high-dimensional weighted gene co-expression network analysis, conversion of features into two-dimensional disease maps for CNN classification, and molecular docking, applied here to liver cirrhosis. It reports an endothelial subpopulation plus seven signature genes and states that the CNN step beat conventional methods. The framework is presented as reusable for other omics settings with noise and small samples. That integration is the actual new piece; the individual tools are not novel on their own, but putting them in this sequence for a concrete disease case gives a workable template. The approach addresses real pain points like feature correlations and limited labels by trying to stabilize modules and capture non-linear effects through the 2D reshaping. Those choices are reasonable on their face and show clear thinking about the data characteristics. The soft spots sit right where the stress-test note flags them. No numbers appear for classification accuracy, no error bars, no validation splits, no baseline tables, and no ablation showing that the 2D map step or the hdWGCNA parameters actually drive the reported gains. Without evidence that the network modules remain stable under the sparsity and noise typical of single-cell data, or that the tabular-to-2D conversion improves results over direct tabular models, the seven-gene list and the outperformance statement rest on untested assumptions. If the full methods section contains those checks and the code is shared, the paper strengthens considerably; as described, the central results cannot be evaluated. This work is for bioinformaticians and applied ML researchers who build pipelines for specific diseases and want an example of how to sequence these steps. A reader looking for a starting scaffold rather than a finished method could extract useful structure. It deserves peer review because the overall framing is coherent and the application is grounded in a real clinical problem, even though revisions would need to add the missing quantitative validation and sensitivity tests before any stronger claims could stand.

Referee Report

3 major / 2 minor

Summary. The manuscript proposes a multi-stage soft computing framework for complex disease modeling and decision support, demonstrated on liver cirrhosis using single-cell transcriptomics. It combines identification of cellular subpopulations, high-dimensional weighted gene co-expression network analysis (hdWGCNA) to stabilize gene modules, restructuring of tabular features into 2D disease maps for CNN-based classification, and molecular docking for therapeutic evaluation. The central claims are the discovery of a disease-associated endothelial subpopulation and seven signature genes (HSPB1, GADD45A, CLDN5, ATP1B3, C1QBP, ENPP2, PARL), superior CNN performance over conventional pipelines, and the framework's disease-agnostic applicability to other omics settings with uncertainty and limited samples.

Significance. If the empirical validations and methodological details hold upon revision, the work could provide a practical, integrated pipeline for extracting interpretable biological signals from noisy, high-dimensional single-cell data while incorporating downstream therapeutic screening. The disease-agnostic framing and emphasis on robustness under sparsity are potentially valuable for the field, though the current lack of quantitative support limits immediate assessment of novelty relative to existing multi-omics frameworks.

major comments (3)

[Abstract] Abstract: the claim that the CNN-based representation learning module outperformed conventional pipelines is unsupported by any quantitative metrics, error bars, validation splits, baseline comparisons, or statistical tests, leaving the central assertion of superiority without evidence.
[Methods] Methods (hdWGCNA description): the assertion that hdWGCNA reliably stabilises gene modules under sparsity and noise lacks any reported stability metrics, hyperparameter settings, module preservation statistics, or ablation on synthetic sparse/noisy data, making it impossible to evaluate whether the seven signature genes follow from this step.
[Methods] Methods (2D disease map construction): no explicit grid-construction rule, dimensionality choice, or ablation study is provided to demonstrate that converting tabular molecular features into 2D maps enables the CNN to capture non-linear interactions better than direct tabular models or standard ML baselines; without this, the reported classification gain cannot be attributed to the proposed transformation.

minor comments (2)

[Abstract] The abstract lists the seven genes without accompanying importance scores, p-values, or cross-validation stability measures that would normally appear in a results summary.
The manuscript would benefit from a dedicated reproducibility subsection listing all software versions, random seeds, and data preprocessing steps for the single-cell analysis pipeline.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed comments on our manuscript. We address each major comment point by point below. Where the comments correctly identify gaps in quantitative support or methodological transparency, we will revise the manuscript to incorporate the requested details, metrics, and ablations.

read point-by-point responses

Referee: [Abstract] Abstract: the claim that the CNN-based representation learning module outperformed conventional pipelines is unsupported by any quantitative metrics, error bars, validation splits, baseline comparisons, or statistical tests, leaving the central assertion of superiority without evidence.

Authors: We agree that the abstract claim requires explicit quantitative support to be fully substantiated. Although the full manuscript reports classification results, we will expand the results section to include comprehensive metrics (accuracy, precision, recall, F1-score), error bars from repeated runs or cross-validation, details on validation splits (e.g., stratified k-fold), direct comparisons to standard baselines (logistic regression, random forest, SVM, XGBoost), and statistical tests (e.g., McNemar's test or paired t-tests). The abstract will be updated to reference these additions. This will provide the necessary evidence for the superiority assertion. revision: yes
Referee: [Methods] Methods (hdWGCNA description): the assertion that hdWGCNA reliably stabilises gene modules under sparsity and noise lacks any reported stability metrics, hyperparameter settings, module preservation statistics, or ablation on synthetic sparse/noisy data, making it impossible to evaluate whether the seven signature genes follow from this step.

Authors: We acknowledge that the current methods description lacks the requested quantitative validation for hdWGCNA stability. In the revision, we will specify all hyperparameter settings (softPower, minModuleSize, etc.), report module preservation statistics (Z-summary scores across bootstrap or perturbed samples), and add an ablation study on synthetic sparse/noisy datasets to demonstrate robustness. This will explicitly link the stabilized modules to the derivation of the seven signature genes (HSPB1, GADD45A, CLDN5, ATP1B3, C1QBP, ENPP2, PARL). revision: yes
Referee: [Methods] Methods (2D disease map construction): no explicit grid-construction rule, dimensionality choice, or ablation study is provided to demonstrate that converting tabular molecular features into 2D maps enables the CNN to capture non-linear interactions better than direct tabular models or standard ML baselines; without this, the reported classification gain cannot be attributed to the proposed transformation.

Authors: We agree that the 2D disease map construction requires more explicit description and validation. We will revise the methods to detail the grid-construction rule (e.g., feature arrangement via correlation-based or clustering-driven mapping to preserve neighborhood structure), the choice of dimensionality (e.g., nearest square grid size based on feature count), and include an ablation study comparing the 2D map + CNN pipeline against direct tabular models (MLP, XGBoost) and standard ML baselines. This will allow attribution of any performance gains to the proposed transformation. revision: yes

Circularity Check

0 steps flagged

No circularity; standard external methods applied without self-referential reduction

full rationale

The manuscript describes a pipeline that applies externally defined techniques (single-cell profiling, hdWGCNA for module stabilization, tabular-to-2D map conversion for CNN, and molecular docking) to cirrhosis scRNA-seq data. No equations, parameter-fitting steps, or derivations are presented that equate outputs to inputs by construction. Claims of an endothelial subpopulation and seven signature genes follow from running these standard tools; the CNN superiority is reported as an empirical outcome rather than a fitted or renamed result. No self-citations serve as load-bearing uniqueness theorems or ansatzes. The derivation chain is therefore independent of the paper's own fitted values or prior author work.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on domain-standard assumptions about single-cell data utility and network stabilization methods without introducing new free parameters, invented entities, or ad-hoc axioms beyond those typical in computational biology.

axioms (2)

domain assumption High-dimensional weighted gene co-expression network analysis (hdWGCNA) can stabilise gene modules under sparsity and noise.
Invoked explicitly for the feature stabilisation stage following single-cell profiling.
domain assumption Restructuring tabular molecular features into two-dimensional disease maps enables convolutional neural networks to model non-linear feature interactions.
Basis for the deep representation construction module described in the framework.

pith-pipeline@v0.9.0 · 5617 in / 1621 out tokens · 42329 ms · 2026-05-08T04:53:58.371273+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

40 extracted references

[1]

Gracia-Sancho, G

J. Gracia-Sancho, G. Marrone, and A. Fernandez-Iglesias. Hepatic microcirculation and mechanisms of portal hypertension.Nat Rev Gastroenterol Hepatol, 16:221–234, 2019. X. Huang et al.:Preprint submitted to ElsevierPage 18 of 20 A multi-stage soft computing framework for liver cirrhosis

2019
[2]

Sarin, M

S.K. Sarin, M. Kumar, M. Eslam, J. George, M. Al Mahtab, S.M.F. Akbar, J. Jia, Q. Tian, R. Aggarwal, D.H. Muljono, M. Omata, Y. Ooka, K.H.Han,H.W.Lee,W.Jafri,A.S.Butt,C.H.Chong,S.G.Lim,R.F.Pwu,andD.S.Chen. LiverdiseasesintheAsia-Pacificregion:aLancet Gastroenterology & Hepatology commission.Lancet Gastroenterol Hepatol, 5:167–228, 2020

2020
[3]

Libbrecht and W.S

M.W. Libbrecht and W.S. Noble. Machine learning applications in genetics and genomics.Nat Rev Genet, 16:321–332, 2015

2015
[4]

Aprimerondeeplearningingenomics.NatGenet,51:12–18,2019

J.Zou,M.Huss,A.Abid,P.Mohammadi,A.Torkamani,andA.Telenti. Aprimerondeeplearningingenomics.NatGenet,51:12–18,2019

2019
[5]

Higashi, S.L

T. Higashi, S.L. Friedman, and Y. Hoshida. Hepatic stellate cells as key target in liver fibrosis.Adv Drug Deliv Rev, 121:27–42, 2017

2017
[6]

Iwakiri and J

Y. Iwakiri and J. Trebicka. Portal hypertension in cirrhosis: Pathophysiological mechanisms and therapy.JHEP Rep, 3:100316, 2021

2021
[7]

Single-celltranscriptomicsrevealszone-specificalterations of liver sinusoidal endothelial cells in cirrhosis.Cell Mol Gastroenterol Hepatol, 11:1139–1161, 2021

T.Su,Y.Yang,S.Lai,J.Jeong,Y.Jung,M.McConnell,T.Utsumi,andY.Iwakiri. Single-celltranscriptomicsrevealszone-specificalterations of liver sinusoidal endothelial cells in cirrhosis.Cell Mol Gastroenterol Hepatol, 11:1139–1161, 2021

2021
[8]

X. Wu, L. Shu, Z. Zhang, J. Li, J. Zong, L.Y. Cheong, D. Ye, K.S.L. Lam, E. Song, C. Wang, A. Xu, and R.L.C. Hoo. Adipocyte fatty acid binding protein promotes the onset and progression of liver fibrosis via mediating the crosstalk between liver sinusoidal endothelial cells and hepatic stellate cells.Adv Sci (Weinh), 8:e2003721, 2021

2021
[9]

Poisson, S

J. Poisson, S. Lemoinne, C. Boulanger, F. Durand, R. Moreau, D. Valla, and P.E. Rautou. Liver sinusoidal endothelial cells: Physiology and role in liver diseases.J Hepatol, 66:212–227, 2017

2017
[10]

Ramachandran, R

P. Ramachandran, R. Dobie, J.R. Wilson-Kanamori, E.F. Dora, B.E.P. Henderson, N.T. Luu, J.R. Portman, K.P. Matchett, M. Brice, J.A. Marwick,R.S.Taylor,M.Efremova,R.Vento-Tormo,N.O.Carragher,T.J.Kendall,J.A.Fallowfield,E.M.Harrison,D.J.Mole,S.J.Wigmore, P.N. Newsome,C.J. Weston, J.P.Iredale, F.Tacke, J.W. Pollard,C.P. Ponting, J.C.Marioni, S.A.Teichmann, a...

2019
[11]

Luecken and F.J

M.D. Luecken and F.J. Theis. Current best practices in single-cell RNA-seq analysis: a tutorial.Mol Syst Biol, 15:e8746, 2019

2019
[12]

Tutorial:guidelinesforthecomputationalanalysisofsingle-cellRNAsequencing data.Nat Protoc, 16:1–9, 2021

T.S.Andrews,V.Y.Kiselev,D.McCarthy,andM.Hemberg. Tutorial:guidelinesforthecomputationalanalysisofsingle-cellRNAsequencing data.Nat Protoc, 16:1–9, 2021

2021
[13]

Morabito, F

S. Morabito, F. Reese, N. Rahimzadeh, E. Miyoshi, and V. Swarup. hdWGCNA identifies co-expression networks in high-dimensional transcriptomics data.Cell Rep Methods, 3:100498, 2023

2023
[14]

Yasaka, H

K. Yasaka, H. Akai, O. Abe, and S. Kiryu. Deep learning with convolutional neural network for differentiation of liver masses at dynamic contrast-enhanced CT: A preliminary study.Radiology, 286:887–896, 2018

2018
[15]

Y. Xie, H. Shi, and B. Han. Bioinformatic analysis of underlying mechanisms of Kawasaki disease via weighted gene correlation network analysis (WGCNA) and the least absolute shrinkage and selection operator method (LASSO) regression model.BMC Pediatr, 23:90, 2023

2023
[16]

DeepInsight:Amethodologytotransformanon-imagedatatoanimage for convolution neural network architecture.Sci Rep, 9:11399, 2019

A.Sharma,E.Vans,D.Shigemizu,K.A.Boroevich,andT.Tsunoda. DeepInsight:Amethodologytotransformanon-imagedatatoanimage for convolution neural network architecture.Sci Rep, 9:11399, 2019

2019
[17]

Newman, C.L

A.M. Newman, C.L. Liu, M.R. Green, A.J. Gentles, W. Feng, Y. Xu, C.D. Hoang, M. Diehn, and A.A. Alizadeh. Robust enumeration of cell subsets from tissue expression profiles.Nat Methods, 12:453–457, 2015

2015
[18]

Dockingandscoringinvirtualscreeningfordrugdiscovery:methodsandapplications

D.B.Kitchen,H.Decornez,J.R.Furr,andJ.Bajorath. Dockingandscoringinvirtualscreeningfordrugdiscovery:methodsandapplications. Nat Rev Drug Discov, 3:935–949, 2004

2004
[19]

Journal of Hepatology, 78:1216–1233, 2023

M.Bhat,M.Rabindranath,B.S.Chara,andD.A.Simonetto.Artificialintelligence,machinelearning,anddeeplearninginlivertransplantation. Journal of Hepatology, 78:1216–1233, 2023

2023
[20]

Lee, J.J.Y

H.W. Lee, J.J.Y. Sung, and S.H. Ahn. Artificial intelligence in liver disease.Journal of gastroenterology and hepatology, 36:539–542, 2021

2021
[21]

Beam and I.S

A.L. Beam and I.S. Kohane. Big data and machine learning in health care.JAMA, 319:1317–1318, 2018

2018
[22]

Ahn, Z.I

J.C. Ahn, Z.I. Attia, P. Rattan, A.F. Mullan, S. Buryska, A.M. Allen, P.S. Kamath, P.A. Friedman, V.H. Shah, P.A. Noseworthy, and D.A. Simonetto. Development of the AI-Cirrhosis-ECG score: An electrocardiogram-based deep learning model in cirrhosis.The American journal of gastroenterology, 117:424–432, 2022

2022
[23]

Langfelder and S

P. Langfelder and S. Horvath. WGCNA: an R package for weighted correlation network analysis.BMC Bioinformatics, 9:559, 2008

2008
[24]

Sturm, F

G. Sturm, F. Finotello, F. Petitprez, J.D. Zhang, J. Baumbach, W.H. Fridman, M. List, and T. Aneichyk. Comprehensive evaluation of transcriptome-based cell-type quantification methods for immuno-oncology.Bioinformatics, 35:i436–i445, 2019

2019
[25]

Fuzzylogicapplicationsinchronicdiseaseindecision-making:acomprehensivehealthcarereview.Informaticsfor Health and Social Care, pages 1–17, 2025

S.ThukralandA.Gupta. Fuzzylogicapplicationsinchronicdiseaseindecision-making:acomprehensivehealthcarereview.Informaticsfor Health and Social Care, pages 1–17, 2025

2025
[26]

Gadaras and L

I. Gadaras and L. Mikhailov. An interpretable fuzzy rule-based classification methodology for medical diagnosis.Artificial Intelligence in Medicine, 47:25–41, 2009

2009
[27]

Ageneselectionalgorithmformicroarraycancer classification using an improved particle swarm optimization.Scientific Reports, 14:19613, 2024

A.A.Nagra,A.H.Khan,M.Abubakar,M.Faheem,A.Rasool,K.Masood,andM.Hussain. Ageneselectionalgorithmformicroarraycancer classification using an improved particle swarm optimization.Scientific Reports, 14:19613, 2024

2024
[28]

Seera and C.P

M. Seera and C.P. Lim. A hybrid intelligent system for medical data classification.Expert Systems with Applications, 41:2239–2249, 2014

2014
[29]

S. She, M. Yang, H. Hu, P. Hu, Y. Yang, and H. Ren. Proteomics based identification of autotaxin as an anti-hepatitis B virus factor and a promoter of hepatoma cell invasion and migration.Cell Physiol Biochem, 45:744–760, 2018

2018
[30]

Z. Xie, Y. Gao, C. Ho, L. Li, C. Jin, X. Wang, C. Zou, Y. Mao, X. Wang, Q. Li, D. Fu, and Y.F. Zhang. Exosome-delivered CD44v6/C1QBP complex drives pancreatic cancer liver metastasis by promoting fibrotic liver microenvironment.Gut, 71:568–579, 2022

2022
[31]

Hagan, S.K

M. Hagan, S.K. Asrani, and J. Talwalkar. Non-invasive assessment of liver fibrosis and prognosis.Expert Rev Gastroenterol Hepatol, 9: 1251–1260, 2015

2015
[32]

L.D. DeLeve. Liver sinusoidal endothelial cells in hepatic fibrosis.Hepatology, 61:1740–1746, 2015

2015
[33]

Caveolin1-relatedautophagy initiated by aldosterone-induced oxidation promotes liver sinusoidal endothelial cells defenestration.Redox Biol, 13:508–521, 2017

X.Luo,W.Dan,X.Luo,X.Zhu,G.Wang,Z.Ning,Y.Li,X.Ma,R.Yang,S.Jin,Y.Huang,Y.Meng,andX.Li. Caveolin1-relatedautophagy initiated by aldosterone-induced oxidation promotes liver sinusoidal endothelial cells defenestration.Redox Biol, 13:508–521, 2017

2017
[34]

G. Xie, X. Wang, L. Wang, L. Wang, R.D. Atkinson, G.C. Kanel, W.A. Gaarde, and L.D. Deleve. Role of differentiation of liver sinusoidal endothelial cells in progression and regression of hepatic fibrosis in rats.Gastroenterology, 142:918–927.e6, 2012

2012
[35]

Tsuchida and S.L

T. Tsuchida and S.L. Friedman. Mechanisms of hepatic stellate cell activation.Nat Rev Gastroenterol Hepatol, 14:397–411, 2017

2017
[36]

Krenkel and F

O. Krenkel and F. Tacke. Liver macrophages in tissue homeostasis and disease.Nat Rev Immunol, 17:306–321, 2017. X. Huang et al.:Preprint submitted to ElsevierPage 19 of 20 A multi-stage soft computing framework for liver cirrhosis

2017
[37]

Seki and R.F

E. Seki and R.F. Schwabe. Hepatic inflammation and fibrosis: functional links and key pathways.Hepatology, 61:1066–1079, 2015

2015
[38]

EsculininhibitshepaticstellatecellactivationandCCl4-induced liver fibrosis by activating the Nrf2/GPX4 signaling pathway.Phytomedicine, 128:155465, 2024

S.Xu,Y.Chen,J.Miao,Y.Li,J.Liu,J.Zhang,J.Liang,S.Chen,andS.Hou. EsculininhibitshepaticstellatecellactivationandCCl4-induced liver fibrosis by activating the Nrf2/GPX4 signaling pathway.Phytomedicine, 128:155465, 2024

2024
[39]

Ranalli, N

M.G. Ranalli, N. Salvati, L. Petrella, and F. Pantalone. M-quantile regression shrinkage and selection via the Lasso and Elastic Net to assess the effect of meteorology and traffic on air quality.Biom J, 65:e2100355, 2023

2023
[40]

Forli, R

S. Forli, R. Huey, M.E. Pique, M.F. Sanner, D.S. Goodsell, and A.J. Olson. Computational protein-ligand docking and virtual drug screening with the AutoDock suite.Nat Protoc, 11:905–919, 2016. X. Huang et al.:Preprint submitted to ElsevierPage 20 of 20

2016