Multi-objective evolutionary algorithms based Interpretable Fuzzy models for microarray gene expression data analysis

An approach to construct interpretable and precise fuzzy models from data is proposed. Interpretability, which is one of the most important features of fuzzy models, is analyzed first. Then a modified fuzzy clustering algorithm, combined with the least square method, is used to identify the initial fuzzy model. Third, the multi-objective genetic algorithm and interpretability-driven simplification techniques are proposed to evolve the initial fuzzy model to optimize its structure and parameters iteratively, thus interpretability and precision of the fuzzy model are improved. Finally, the proposed approach is applied to the Mackey-Glass tine series, and the results show its validity

The phosphate starvation response in bacteria has been studied extensively for the past few decades and the phosphate-limiting signal is known to be mediated via the PhoBR two-component system. However, the global DNA binding profile of the response regulator PhoB and the PhoB downstream responses are currently unclear. In this study, chromatin immunoprecipitation for PhoB was combined with high-density tiling array (ChIP-chip) as well as gene expression microarray to reveal the first global down-stream responses of the responding regulator, PhoB in E. coli. Based on our ChIP-chip experimental data, forty-three binding sites were identified throughout the genome and the known PhoB binding pattern was updated by identifying the conserved pattern from these sites. From the gene expression microarray data analysis, 287 differentially expressed genes were identified in the presence of PhoB activity. By comparing the results obtained from our ChIP-chip and microarray experiments, we were also able to identify genes that were directly or indirectly affected through PhoB regulation. Nineteen out of these 287 differentially expressed genes were identified as the genes directly regulated by PhoB. Seven of the 19 directly regulated genes (including phoB) are transcriptional regulators. These transcriptional regulators then further pass the signal of phosphate starvation down to the remaining differentially expressed genes. Our results unveiled the genome-wide binding profile of PhoB and the downstream responses under phosphate starvation. We also present the hierarchical structure of the phosphate sensing regulatory network. The data suggest that PhoB plays protective roles in membrane integrity and oxidative stress reduction during phosphate starvation.

The dynamic nature and increasing complexity of the construction industry have led to increased conflicts in construction projects. An accurate prediction of the outcome of a dispute resolution in courts could effectively reduce the number of disputes that would otherwise conclude by spending more money through litigation. This study aims to introduce a two-layered fuzzy logic model for predicting court decisions in construction contract disputes. 100 cases of construction contract disputes are selected from the courts of Iran. A questionnaire survey is then conducted to extract a set of fuzzy rules for identifying important decision parameters and expert knowledge. Accordingly, a two-layered fuzzy logic-based decision-making architecture is proposed for the prediction model. Furthermore, the fuzzy system is trained based on 10-fold cross-validation. Analysis of results indicates that 51 out of the 100 cases are filed after the dissolution and termination of the contract show a significant impact of these clauses as the root cause in construction contract disputes. Our results present a proposed hierarchical fuzzy system that can correctly predict nearly 60% of the test data. Also, we demonstrate a methodology of using argument before ML to establish interpretable AI models. Based on our findings, a fuzzy model with a hierarchical structure may be used as a simple and efficient method for predicting court decisions in construction contract disputes.

Interpretability improvements to find the balance interpretability-accuracy in fuzzy modeling: an overview.- Regaining comprehensibility of approximative fuzzy models via the use of linguistic hedges.- Identifying flexible structured premises for mining concise fuzzy knowledge.- A multiobjective genetic learning process for joint feature selection and granularity and contexts learning in fuzzy rule-based classification systems.- Extracting linguistic fuzzy models from numerical data-AFRELI algorithm.- Constrained optimization of fuzzy decision trees.- A new method for inducing a set of interpretable fuzzy partitions and fuzzy inference systems from data.- A Feature Ranking Algorithm for Fuzzy Modelling Problems.- Interpretability in multidimensional classification.- Interpretable semi-mechanistic fuzzy models by clustering, OLS and FIS model reduction.- Trade-off between approximation accuracy and complexity: TS controller design via HOSVD based complexity minimization.- Simplification and reduction of fuzzy rules.- Effect of rule representation in rule base reduction.- Singular value-based fuzzy reduction with relaxed normalization condition.- Interpretability, complexity, and modular structure of fuzzy systems.- Hierarchical genetic fuzzy systems: accuracy, interpretability and design autonomy.- About the trade-off between accuracy and interpretability of Takagi-Sugeno models in the context of nonlinear time series forecasting.- Accurate, transparent and compact fuzzy models by multi-objective evolutionary algorithms.- Transparent fuzzy systems in modeling and control.- Uniform fuzzy partitions with cardinal splines and wavelets: getting interpretable linguistic fuzzy models.- Relating the theory of partitions in MV-logic to the design of interpretable fuzzy systems.- A formal model of interpretability of linguistic variables.- Expressing relevance and interpretability of rule-based systems.- Conciseness of fuzzy models.- Exact trade-off between approximation accuracy and interpretability: solving the saturation problem for certain FRBSs.- Interpretability improvement of RBF-based neurofuzzy systems using regularized learning.- Extracting fuzzy classification rules from fuzzy clusters on the basis of separating hyperplanes.

The important purpose of the microarray gene expression data analysis is to identify significantly differentially-expressed genes between two groups of samples which are in two different experimental states. In this work, we propose to use several statistical test methods for a given microarray data set and cross-refer the results of different statistical test methods. The accuracy of different statistical methods is estimated by Receiver Operation Characteristic (ROC) technique. A new software tool, MicroMultitest, was developed. A number of statistical testing methods (such as t-test, adapted SAM method, p-value adjustments), as well as the ROC analysis technique were implemented in this software. Using the MicroMultitest one has the ability to evaluate the performance of different statistical testing methods by applying each to the same given microarray data set, optimize the cutoff values and permutation times for these statistical testing methods, and select relative reliable differentially-expressed gene set.

Rapid technological advancements in microarray analysis continue to generate enormous amounts of genomic data. However, neither hardware nor software computational capabilities have kept pace with this drastic increase. This paper presents a novel framework designed to achieve fast, robust, and accurate (biologically-significant) multi-class classification of gene expression data using distributed knowledge discovery and computational integration routines, specifically for cancer applications. The proposed paradigm consists of the following key computational steps: (a) preprocessing normalization and discretization of gene expression data, (b) partition data using two methods: overlapped windows and adaptive selection, (c) perform association rule discovery on partitioned data-spaces using FP-growth method, (d) integrate derived association rules on distributed processor nodes using a novel knowledge integration algorithm, (e) further prune rules to reduce dimensionality using parametric significance estimation, and (f) cluster remaining rules using a novel clustering algorithm for enhanced visualization and interpretation of discovered gene rule sets.

Cancer disease persists as one of the major causes of mortality worldwide, emphasizing the critical necessity for accurate and early diagnosis. Identifying genes associated with cancer can facilitate effective early-stage treatment. DNA Microarray data plays a vital role in detecting and diagnosing cancer. Microarray analysis allows the examination of gene expression levels in specific cell samples, enabling the simultaneous analysis of thousands of genes. However, microarray gene data is characterized by high dimensionality and contains redundant genes. The gene (feature) selection process is essential in microarray gene expression data analysis for selecting highly relevant genes with low redundancy that may cause improved prediction results. This research article presents a model designed to reduce the high-dimensional gene data to low-dimensional space, eliminate redundant genes, and improve early-stage disease prediction. The proposed model is implemented in the following phases. The initial process involves checking the multicollinearity between genes and identifying the low intercorrelated genes using the Karl Pearson correlation coefficient on the cancer dataset. Next, the Genetic Optimization Algorithm integrated with the Signal to Noise Ratio method is applied to determine an optimal set of genes, focusing on reducing dimensionality while filtering out noisy and redundant genes. In the final phase, the selected genes are classified using a Deep learning classification method namely the Long Short Term Memory classifier is utilized. The model’s performance is evaluated with various optimizers, including Adam, Adagrad, RMSProp, and SGD by comparing accuracy and loss values. Also calculate the other classification parameters such as Precision, Recall, and F1-Score value. Finally, comparative analyses of these metrics across the different optimizers are performed. The experimental results demonstrate that the research model Cor-GA_SNR-LSTM significantly improves the detection analysis by reducing the number of features in the gene data, enhancing the accuracy value, and minimizing the loss value. The Adam optimizer yielded the optimal performance with an accuracy of 88.45% and a loss value of 0.1612. The proposed method effectively identifies the most relevant genes responsible for detecting and predicting Leukemia cancer disease.

In the rapidly advancing field of genomics, microarray technologies have turned into a ground-breaking system on simultaneous monitoring the expression patterns of multiple genes under various arrangements of constraints. A fundamental errand is to propose diagnostic techniques to distinguish cluster of genes comparative expression designs and are initiated by comparative conditions. And furthermore, the relating investigation has issue is to cluster multi-condition gene expression data. To overcome these issues, the vast measure of data obtained by this technology, resort to clustering methods that distinguish clusters of genes of share similar expression profiles. The motivation of this work is to introduce a clustering method in microarray gene expression data analysis. Multi-Objective Binary Particle Swarm Optimization with Fuzzy Weighted Clustering (MOBPSOFWC) algorithm is proposed to analyze gene expression data. In high dimensionality, a quick heuristic based pre-processing technique is employed to diminish some of the basic domain features from the initial feature set. Since these pre-processed and reduced features are still high dimensional, the proposed MOBPSO algorithm is implemented in MATLAB tool used for finding further feature subsets. The investigative are directed to distinguish the execution of the proposed work with existing clustering approaches.

Clustering is one of most useful tools for the microarray gene expression data analysis. Although there have been many reviews and surveys in the literature, many good and effective clustering ideas have not been collected in a systematic way for some reasons. In this paper, we review five clustering families representing five clustering concepts rather than five algorithms. We also review some clustering validations and collect a list of benchmark gene expression datasets.

Gridding and spot segmentation are two critical steps in microarray gene expression data analysis. However, the problems of noise contamination and donut-shaped spots often make signal extraction process a laborintensive task. In this paper, we propose a three-step method for automatic gridding and spot segmentation. The method starts with a background removal and noise eliminating step, and then proceeds in two steps. The first step applies a fully unsupervised method to extract blocks and grids from the cleaned data. The second step applies a simple, progressive spot segmentation method to deal with inner holes and noise in spots. We tested its performance on real microarray images against a widely used software GenePix. Our results show that the proposed method deals effectively with poor-conditioned microarray images in both gridding and spot segmentation.

Analyzing microarray gene expression data provides biologists deep insights into gene functions and gene regulatory network. In this paper, we propose a novel efficient algorithm FCPminer to mine top-k frequent closed patterns (FCPs) of higher support with length no less than minL from gene expression data. FCPminer employs a prefix fp-tree data structure, with top-down best first search strategy, such that FCPs of adequate length with highest supports are firstly mined. Compared with existing top-k FCP mining algorithms, FCPminer is much more efficient as it avoids expanding nodes with inadequate length (less than minL) or low support (ranked below top-k) during mining process. In addition, FCPminer further improves mining efficiency by employing a hash-based closedness checking method. Experimental results on real biological and synthetic data show that our proposed FCPminer outperforms existing state-of the art algorithms with high efficiency, especially for large and dense datasets.

CD47 Is An Independent Prognostic Factor and Therapeutic Antibody Target on Human Acute Myeloid Leukemia Stem Cells

Background: To assess recurrence risk in ER-positive, HER2-negative early breast cancer, multigene assays (MGAs) are useful in addition to clinicopathological factors. In Japan, Curebest™ 95GC was developed in 2011 and introduced clinically in 2013. Oncotype DX® was added to national health insurance in 2023. While adjuvant chemotherapy is often considered for patients classified as high-risk by any MGA, treatment decisions can be complex. Furthermore, breast cancers identified as high-risk by MGAs often exhibit higher pathological complete response (pCR) rates, suggesting increased sensitivity to chemotherapy. This study aimed to evaluate MGAs from two perspectives—prognostic prediction and NAC sensitivity—to inform individualized treatment strategies. Materials and Methods: We analyzed microarray gene expression data from public databases to identify patients with ER-positive, HER2-negative early breast cancer. In this study, Curebest™ 95GC, developed in Japan, was used post-analysis for research purposes only. Based on microarray expression data, we calculated PAM50, 21GC (Oncotype DX® surrogate analyses), EndoPredict (EP), and either 95GC or 155GC. In Study 1(n=555), we assessed the prognostic value of each MGA and evaluated recurrence risk based on the number of high-risk classifications. In study2(n=550), we evaluated the predictive value of each MGA for pCR rate and analyzed the relationship between NAC sensitivity and the combined assessment of the number of high-risk classifications treated with NAC. Results: In Study 1, all four MGAs (PAM50, 21GC, EP, and 95GC) significantly stratified DRFS between high- and low-risk groups (p =8.93e-08, 2.31e-08, 3.64e-07, and 9.25e-10), with comparable prognostic performance. Patients were grouped by the number of high-risk classifications. A 5-year DRFS of 97% was observed in patients with 0 high-risk classifications, compared to 66% in those with 4. Patients were further stratified into two groups: those with 0 or 1 high-risk classification and those with ≥2. The 0-1 group showed a 5-year DRFS of 96.7%, while the ≥2 group had a significantly worse prognosis at 79.0% (p=2.04e-11). The combined assessment provided better prognostic accuracy than any individual MGA alone. Notably, patients with only one high-risk result showed excellent outcomes, suggesting endocrine therapy alone may suffice. Conversely, those deemed low-risk by all MGAs may not require adjuvant chemotherapy.In Study 2, high-risk patients had significantly higher pCR rates across all MGAs. 155GC showed the strongest stratification (p=1.62e-08). Patients classified as high-risk by all four MGAs had significantly higher pCR rates than those with 0-3 high-risk classifications (p=9.37e-11). 95GC provides CEL files containing expression data for all 20,000 genes. This allows concurrent calculation of multiple MGAs, facilitating evaluation and reducing cost. Use of CEL files supports integrated MGA analysis, and may support precise personalized treatment. Conclusion: Combining four MGAs improves prediction over single MGAs. This integrated genomic approach offers a promising strategy to optimize individualized treatment in ER-positive, HER2-negative early breast cancer. Citation Format: M. Nishida, R. Tsunashima, N. Hirotani, S. Kitano, C. Matsui, S. Matsumoto, A. Watanabe, C. Kato, M. Morita, K. Sakaguchi, Y. Naoi. Integrated Prognostic and Chemotherapy Sensitivity Assessment Using Multiple Gene Assays in ER-positive, HER2-negative Early Breast Cancer [abstract]. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS3-08-25.

Although gene expression microarray data analysis has been developed and used during the last 15 years, microRNA (miRNA) analysis is still under development and important aspects affecting the final results such as normalization still remain unclear. In this work we have studied and compared the effect of non normalization and four different normalization strategies (75th percentile, hsa-mir103 and hsa-mir-let7a control based and median of all expressed miRNA) on a data set of colon cancer miRNA microarrays. Different subsets of samples were used to study normalization effects comparing data distributions before and after normalization with the different strategies. Median of expressed miRNA behaved the best among the normalizations studied. Class comparison based analysis showed that differentially expressed miRNA are highly dependent on the normalization techniques applied.

Fuzzy models occupy one of the dominant positions on the research agenda of fuzzy sets exhibiting a wealth of conceptual developments and algorithmic pursuits as well as a plethora of applications. Granular fuzzy modeling dwelling on the principles of fuzzy modeling opens new horizons of investigations and augments the existing design methodology exploited in fuzzy modeling. In a nutshell, granular fuzzy models are constructs built upon fuzzy models or a family of fuzzy models. We elaborate on a number of compelling reasons behind the emergence of granular fuzzy modelling, and granular modeling, in general. Information granularity present in such models plays an important role. Given a fuzzy model M, the associated granular model incorporates granular information to quantify a performance of the original model, facilitate collaborative pursuits of knowledge management and knowledge transfer. We discuss several main categories of granular fuzzy models where such categories depend upon the formalism of information granularity giving rise to interval-valued fuzzy models, fuzzy fuzzy model (fuzzy2 models, for short), and rough -fuzzy models. The design of granular fuzzy models builds upon two fundamental concepts of Granular Computing: the principle of justifiable granularity and an optimal allocation (distribution) of information granularity. The first one supports a construction of information granules of a granular fuzzy model. The second one emphasizes the role of information granularity being treated as an important design asset. The underlying performance indexes guiding the design of granular fuzzy models are discussed and a multiobjective nature of the construction of these models is stressed.Keywordsinformation granularitygranular fuzzy modelsprinciple of justifiable granularityknowledge managementgranularity allocation