A survey of neural network-based cancer prediction models from microarray data

Machine learning for the advancement of genome-scale metabolic modeling

Constraint-based modeling (CBM) has evolved as the core systems biology tool to map the interrelations between genotype, phenotype, and external environment. The recent advancement of high-throughput experimental approaches and multi-omics strategies has generated a plethora of new and precise information from wide-ranging biological domains. On the other hand, the continuously growing field of machine learning (ML) and its specialized branch of deep learning (DL) provide essential computational architectures for decoding complex and heterogeneous biological data. In recent years, both multi-omics and ML have assisted in the escalation of CBM. Condition-specific omics data, such as transcriptomics and proteomics, helped contextualize the model prediction while analyzing a particular phenotypic signature. At the same time, the advanced ML tools have eased the model reconstruction and analysis to increase the accuracy and prediction power. However, the development of these multi-disciplinary methodological frameworks mainly occurs independently, which limits the concatenation of biological knowledge from different domains. Hence, we have reviewed the potential of integrating multi-disciplinary tools and strategies from various fields, such as synthetic biology, CBM, omics, and ML, to explore the biochemical phenomenon beyond the conventional biological dogma. How the integrative knowledge of these intersected domains has improved bioengineering and biomedical applications has also been highlighted. We categorically explained the conventional genome-scale metabolic model (GEM) reconstruction tools and their improvement strategies through ML paradigms. Further, the crucial role of ML and DL in omics data restructuring for GEM development has also been briefly discussed. Finally, the case-study-based assessment of the state-of-the-art method for improving biomedical and metabolic engineering strategies has been elaborated. Therefore, this review demonstrates how integrating experimental and in silico strategies can help map the ever-expanding knowledge of biological systems driven by condition-specific cellular information. This multiview approach will elevate the application of ML-based CBM in the biomedical and bioengineering fields for the betterment of society and the environment.

Machine learning models for predicting of PD-1 treatment efficacy in Pan-cancer patients based on routine hematologic and biochemical parameters

Immune checkpoint blockade therapy targeting the programmed death-1(PD-1) pathway has shown remarkable efficacy and durable response in patients with various cancer types. Early prediction of therapeutic efficacy is important for optimizing treatment plans and avoiding potential side effects. In this work, we developed an efficient machine learning prediction method using routine hematologic and biochemical parameters to predict the efficacy of PD-1 combination treatment in Pan-Cancer patients. A total of 431 patients with nasopharyngeal carcinoma, esophageal cancer and lung cancer who underwent PD-1 checkpoint inhibitor combination therapy were included in this study. Patients were divided into two groups: progressive disease (PD) and disease control (DC) groups. Hematologic and biochemical parameters were collected before and at the third week of PD-1 therapy. Six machine learning models were developed and trained to predict the efficacy of PD-1 combination therapy at 8-12 weeks. Analysis of 57 blood biomarkers before and after three weeks of PD-1 combination therapy through statistical analysis, heatmaps, and principal component analysis did not accurately predict treatment outcome. However, with machine learning models, both the AdaBoost classifier and GBDT demonstrated high levels of prediction efficiency, with clinically acceptable AUC values exceeding 0.7. The AdaBoost classifier exhibited the highest performance among the 6 machine learning models, with a sensitivity of 0.85 and a specificity of 0.79. Our study demonstrated the potential of machine learning to predict the efficacy of PD-1 combination therapy based on changes in hematologic and biochemical parameters.

Unveiling the Future of Oncology and Precision Medicine through Data Science

Information generated via next-generation sequencing (NGS) technologies is often termed multi-omics data [...].

Optimized model architectures for deep learning on genomic data

The success of deep learning in various applications depends on task-specific architecture design choices, including the types, hyperparameters, and number of layers. In computational biology, there is no consensus on the optimal architecture design, and decisions are often made using insights from more well-established fields such as computer vision. These may not consider the domain-specific characteristics of genome sequences, potentially limiting performance. Here, we present GenomeNet-Architect, a neural architecture design framework that automatically optimizes deep learning models for genome sequence data. It optimizes the overall layout of the architecture, with a search space specifically designed for genomics. Additionally, it optimizes hyperparameters of individual layers and the model training procedure. On a viral classification task, GenomeNet-Architect reduced the read-level misclassification rate by 19%, with 67% faster inference and 83% fewer parameters, and achieved similar contig-level accuracy with ~100 times fewer parameters compared to the best-performing deep learning baselines.

Optimized models and deep learning methods for drug response prediction in cancer treatments: a review

Recent advancements in deep learning (DL) have played a crucial role in aiding experts to develop personalized healthcare services, particularly in drug response prediction (DRP) for cancer patients. The DL's techniques contribution to this field is significant, and they have proven indispensable in the medical field. This review aims to analyze the diverse effectiveness of various DL models in making these predictions, drawing on research published from 2017 to 2023. We utilized the VOS-Viewer 1.6.18 software to create a word cloud from the titles and abstracts of the selected studies. This study offers insights into the focus areas within DL models used for drug response. The word cloud revealed a strong link between certain keywords and grouped themes, highlighting terms such as deep learning, machine learning, precision medicine, precision oncology, drug response prediction, and personalized medicine. In order to achieve an advance in DRP using DL, the researchers need to work on enhancing the models' generalizability and interoperability. It is also crucial to develop models that not only accurately represent various architectures but also simplify these architectures, balancing the complexity with the predictive capabilities. In the future, researchers should try to combine methods that make DL models easier to understand; this will make DRP reviews more open and help doctors trust the decisions made by DL models in cancer DRP.

Deep Learning Techniques to Characterize the RPS28P7 Pseudogene and the Metazoa-SRP Gene as Drug Potential Targets in Pancreatic Cancer Patients

The molecular explanation about why some pancreatic cancer (PaCa) patients die early and others die later is poorly understood. This study aimed to discover potential novel markers and drug targets that could be useful to stratify and extend expected survival in prospective early-death patients. We deployed a deep learning algorithm and analyzed the gene copy number, gene expression, and protein expression data of death versus alive PaCa patients from the GDC cohort. The genes with higher relative amplification (copy number >4 times in the dead compared with the alive group) were EWSR1, FLT3, GPC3, HIF1A, HLF, and MEN1. The most highly up-regulated genes (>8.5-fold change) in the death group were RPL30, RPL37, RPS28P7, RPS11, Metazoa_SRP, CAPNS1, FN1, H3-3B, LCN2, and OAZ1. None of their corresponding proteins were up or down-regulated in the death group. The mRNA of the RPS28P7 pseudogene could act as ceRNA sponging the miRNA that was originally directed to the parental gene RPS28. We propose RPS28P7 mRNA as the most druggable target that can be modulated with small molecules or the RNA technology approach. These markers could be added as criteria to patient stratification in future PaCa drug trials, but further validation in the target populations is encouraged.

Deep Learning Techniques to Characterize the RPS28P7 Pseudogene and the Metazoa-SRP Gene as Drug Potential Targets in Pancreatic Cancer Patients

The molecular explanation about why some pancreatic cancer (PaCa) patients die early and others die later is poorly understood. This study aimed to discover potential novel markers and drug targets that could be useful to stratify and extend expected survival in prospective early-death patients. We deployed a deep learning algorithm and analyzed the gene copy number, gene expression, and protein expression data of death versus alive PaCa patients from the GDC cohort. The genes with higher relative amplification (copy number >4 times in the dead compared with the alive group) were EWSR1, FLT3, GPC3, HIF1A, HLF, and MEN1. The most highly up-regulated genes (>8.5-fold change) in the death group were RPL30, RPL37, RPS28P7, RPS11, Metazoa_SRP, CAPNS1, FN1, H3−3B, LCN2, and OAZ1. None of their corresponding proteins were up or down-regulated in the death group. The mRNA of the RPS28P7 pseudogene could act as ceRNA sponging the miRNA that was originally directed to the parental gene RPS28. We propose RPS28P7 mRNA as the most druggable target that can be modulated with small molecules or the RNA technology approach. These markers could be added as criteria to patient stratification in future PaCa drug trials, but further validation in the target populations is encouraged.

A novel and innovative cancer classification framework through a consecutive utilization of hybrid feature selection

Cancer prediction in the early stage is a topic of major interest in medicine since it allows accurate and efficient actions for successful medical treatments of cancer. Mostly cancer datasets contain various gene expression levels as features with less samples, so firstly there is a need to eliminate similar features to permit faster convergence rate of classification algorithms. These features (genes) enable us to identify cancer disease, choose the best prescription to prevent cancer and discover deviations amid different techniques. To resolve this problem, we proposed a hybrid novel technique CSSMO-based gene selection for cancer classification. First, we made alteration of the fitness of spider monkey optimization (SMO) with cuckoo search algorithm (CSA) algorithm viz., CSSMO for feature selection, which helps to combine the benefit of both metaheuristic algorithms to discover a subset of genes which helps to predict a cancer disease in early stage. Further, to enhance the accuracy of the CSSMO algorithm, we choose a cleaning process, minimum redundancy maximum relevance (mRMR) to lessen the gene expression of cancer datasets. Next, these subsets of genes are classified using deep learning (DL) to identify different groups or classes related to a particular cancer disease. Eight different benchmark microarray gene expression datasets of cancer have been utilized to analyze the performance of the proposed approach with different evaluation matrix such as recall, precision, F1-score, and confusion matrix. The proposed gene selection method with DL achieves much better classification accuracy than other existing DL and machine learning classification models with all large gene expression dataset of cancer.

Classification of cancer cells and gene selection based on microarray data using MOPSO algorithm

PURPOSE

Microarray information is crucial for the identification and categorisation of malignant tissues. The very limited sample size in the microarray has always been a challenge for classification design in cancer research. As a result, by pre-processing gene selection approaches and genes lacking their information, the microarray data are deleted prior to categorisation. In essence, an appropriate gene selection technique can significantly increase the accuracy of illness (cancer) classification.

METHODS

For the classification of high-dimensional microarray data, a novel approach based on the hybrid model of multi-objective particle swarm optimisation (MOPSO) is proposed in this research. First, a binary vector representing each particle's position is presented at random. A gene is represented by each bit. Bit 0 denotes the absence of selection of the characteristic (gene) corresponding to it, while bit 1 denotes the selection of the gene. Therefore, the position of each particle represents a set of genes, and the linear Bayesian discriminant analysis classification algorithm calculates each particle's degree of fitness to assess the quality of the gene set that particle has chosen. The suggested methodology is applied to four different cancer database sets, and the results are contrasted with those of other approaches currently in use.

RESULTS

The proposed algorithm has been applied on four sets of cancer database and its results have been compared with other existing methods. The results of the implementation show that the improvement of classification accuracy in the proposed algorithm compared to other methods for four sets of databases is 25.84% on average. So that it has improved by 18.63% in the blood cancer database, 24.25% in the lung cancer database, 27.73% in the breast cancer database, and 32.80% in the prostate cancer database. Therefore, the proposed algorithm is able to identify a small set of genes containing information in a way choose to increase the classification accuracy.

CONCLUSION

Our proposed solution is used for data classification, which also improves classification accuracy. This is possible because the MOPSO model removes redundancy and reduces the number of redundant and redundant genes by considering how genes are correlated with each other.

Enhancing the prediction of IDC breast cancer staging from gene expression profiles using hybrid feature selection methods and deep learning architecture

Prediction of the stage of cancer plays an important role in planning the course of treatment and has been largely reliant on imaging tools which do not capture molecular events that cause cancer progression. Gene-expression data-based analyses are able to identify these events, allowing RNA-sequence and microarray cancer data to be used for cancer analyses. Breast cancer is the most common cancer worldwide, and is classified into four stages - stages 1, 2, 3, and 4 [2]. While machine learning models have previously been explored to perform stage classification with limited success, multi-class stage classification has not had significant progress. There is a need for improved multi-class classification models, such as by investigating deep learning models. Gene-expression-based cancer data is characterised by the small size of available datasets, class imbalance, and high dimensionality. Class balancing methods must be applied to the dataset. Since all the genes are not necessary for stage prediction, retaining only the necessary genes can improve classification accuracy. The breast cancer samples are to be classified into 4 classes of stages 1 to 4. Invasive ductal carcinoma breast cancer samples are obtained from The Cancer Genome Atlas (TCGA) and Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) datasets and combined. Two class balancing techniques are explored, synthetic minority oversampling technique (SMOTE) and SMOTE followed by random undersampling. A hybrid feature selection pipeline is proposed, with three pipelines explored involving combinations of filter and embedded feature selection methods: Pipeline 1 - minimum-redundancy maximum-relevancy (mRMR) and correlation feature selection (CFS), Pipeline 2 - mRMR, mutual information (MI) and CFS, and Pipeline 3 - mRMR and support vector machine-recursive feature elimination (SVM-RFE). The classification is done using deep learning models, namely deep neural network, convolutional neural network, recurrent neural network, a modified deep neural network, and an AutoKeras generated model. Classification performance post class-balancing and various feature selection techniques show marked improvement over classification prior to feature selection. The best multiclass classification was found to be by a deep neural network post SMOTE and random undersampling, and feature selection using mRMR and recursive feature elimination, with a Cohen-Kappa score of 0.303 and a classification accuracy of 53.1%. For binary classification into early and late-stage cancer, the best performance is obtained by a modified deep neural network (DNN) post SMOTE and random undersampling, and feature selection using mRMR and recursive feature elimination, with an accuracy of 81.0% and a Cohen-Kappa score (CKS) of 0.280. This pipeline also showed improved multiclass classification performance on neuroblastoma cancer data, with a best area under the receiver operating characteristic (auROC) curve score of 0.872, as compared to 0.71 obtained in previous work, an improvement of 22.81%. The results and analysis reveal that feature selection techniques play a vital role in gene-expression data-based classification, and the proposed hybrid feature selection pipeline improves classification performance. Multi-class classification is possible using deep learning models, though further improvement particularly in late-stage classification is necessary and should be explored further.

Genes selection using deep learning and explainable artificial intelligence for chronic lymphocytic leukemia predicting the need and time to therapy

Analyzing gene expression profiles (GEP) through artificial intelligence provides meaningful insight into cancer disease. This study introduces DeepSHAP Autoencoder Filter for Genes Selection (DSAF-GS), a novel deep learning and explainable artificial intelligence-based approach for feature selection in genomics-scale data. DSAF-GS exploits the autoencoder's reconstruction capabilities without changing the original feature space, enhancing the interpretation of the results. Explainable artificial intelligence is then used to select the informative genes for chronic lymphocytic leukemia prognosis of 217 cases from a GEP database comprising roughly 20,000 genes. The model for prognosis prediction achieved an accuracy of 86.4%, a sensitivity of 85.0%, and a specificity of 87.5%. According to the proposed approach, predictions were strongly influenced by CEACAM19 and PIGP, moderately influenced by MKL1 and GNE, and poorly influenced by other genes. The 10 most influential genes were selected for further analysis. Among them, FADD, FIBP, FIBP, GNE, IGF1R, MKL1, PIGP, and SLC39A6 were identified in the Reactome pathway database as involved in signal transduction, transcription, protein metabolism, immune system, cell cycle, and apoptosis. Moreover, according to the network model of the 3D protein-protein interaction (PPI) explored using the NetworkAnalyst tool, FADD, FIBP, IGF1R, QTRT1, GNE, SLC39A6, and MKL1 appear coupled into a complex network. Finally, all 10 selected genes showed a predictive power on time to first treatment (TTFT) in univariate analyses on a basic prognostic model including IGHV mutational status, del(11q) and del(17p), NOTCH1 mutations, β2-microglobulin, Rai stage, and B-lymphocytosis known to predict TTFT in CLL. However, only IGF1R [hazard ratio (HR) 1.41, 95% CI 1.08-1.84, P=0.013), COL28A1 (HR 0.32, 95% CI 0.10-0.97, P=0.045), and QTRT1 (HR 7.73, 95% CI 2.48-24.04, P<0.001) genes were significantly associated with TTFT in multivariable analyses when combined with the prognostic factors of the basic model, ultimately increasing the Harrell's c-index and the explained variation to 78.6% (versus 76.5% of the basic prognostic model) and 52.6% (versus 42.2% of the basic prognostic model), respectively. Also, the goodness of model fit was enhanced (χ2 = 20.1, P=0.002), indicating its improved performance above the basic prognostic model. In conclusion, DSAF-GS identified a group of significant genes for CLL prognosis, suggesting future directions for bio-molecular research.

Unraveling the Molecular Puzzle: Exploring Gene Networks across Diverse EMT Status of Cell Lines

Understanding complex disease mechanisms requires a comprehensive understanding of the gene regulatory networks, as complex diseases are often characterized by the dysregulation and dysfunction of molecular networks, rather than abnormalities in single genes. Specifically, the exploration of cell line-specific gene networks can provide essential clues for precision medicine, as this methodology can uncover molecular interplays specific to particular cell line statuses, such as drug sensitivity, cancer progression, etc. In this article, we provide a comprehensive review of computational strategies for cell line-specific gene network analysis: (1) cell line-specific gene regulatory network estimation and analysis of gene networks under varying epithelial-mesenchymal transition (EMT) statuses of cell lines; and (2) an explainable artificial intelligence approach for interpreting the estimated massive multiple EMT-status-specific gene networks. The objective of this review is to help readers grasp the concept of computational network biology, which holds significant implications for precision medicine by offering crucial clues.

Methodology to identify a gene expression signature by merging microarray datasets

A vast number of microarray datasets have been produced as a way to identify differentially expressed genes and gene expression signatures. A better understanding of these biological processes can help in the diagnosis and prognosis of diseases, as well as in the therapeutic response to drugs. However, most of the available datasets are composed of a reduced number of samples, leading to low statistical, predictive and generalization power. One way to overcome this problem is by merging several microarray datasets into a single dataset, which is typically a challenging task. Statistical methods or supervised machine learning algorithms are usually used to determine gene expression signatures. Nevertheless, statistical methods require an arbitrary threshold to be defined, and supervised machine learning methods can be ineffective when applied to high-dimensional datasets like microarrays. We propose a methodology to identify gene expression signatures by merging microarray datasets. This methodology uses statistical methods to obtain several sets of differentially expressed genes and uses supervised machine learning algorithms to select the gene expression signature. This methodology was validated using two distinct research applications: one using heart failure and the other using autism spectrum disorder microarray datasets. For the first, we obtained a gene expression signature composed of 117 genes, with a classification accuracy of approximately 98%. For the second use case, we obtained a gene expression signature composed of 79 genes, with a classification accuracy of approximately 82%. This methodology was implemented in R language and is available, under the MIT licence, at https://github.com/bioinformatics-ua/MicroGES.

Genetic algorithm-based feature selection with manifold learning for cancer classification using microarray data

BACKGROUND

Microarray data have been widely utilized for cancer classification. The main characteristic of microarray data is "large p and small n" in that data contain a small number of subjects but a large number of genes. It may affect the validity of the classification. Thus, there is a pressing demand of techniques able to select genes relevant to cancer classification.

RESULTS

This study proposed a novel feature (gene) selection method, Iso-GA, for cancer classification. Iso-GA hybrids the manifold learning algorithm, Isomap, in the genetic algorithm (GA) to account for the latent nonlinear structure of the gene expression in the microarray data. The Davies-Bouldin index is adopted to evaluate the candidate solutions in Isomap and to avoid the classifier dependency problem. Additionally, a probability-based framework is introduced to reduce the possibility of genes being randomly selected by GA. The performance of Iso-GA was evaluated on eight benchmark microarray datasets of cancers. Iso-GA outperformed other benchmarking gene selection methods, leading to good classification accuracy with fewer critical genes selected.

CONCLUSIONS

The proposed Iso-GA method can effectively select fewer but critical genes from microarray data to achieve competitive classification performance.

Gene Regulatory Network-Classifier: Gene Regulatory Network-Based Classifier and Its Applications to Gastric Cancer Drug (5-Fluorouracil) Marker Identification

The complex mechanisms of diseases involve the disturbance of the molecular network, rather than disorder in a single gene, implying that single gene-based analysis is insufficient to understand these mechanisms. Gene regulatory networks (GRNs) have attracted a lot of interest and various approaches have been developed for their statistical inference and gene network-based analysis. Although various computational methods have been developed, relatively little attention has been paid to incorporation of biological knowledge into the computational approaches. Furthermore, existing studies on network-based analysis perform prediction/classification of status of cell lines based on preconstructed GRNs, implying that we cannot extract prediction/classification-specific gene networks, leading to difficulty in interpretation of biological mechanisms and marker identification related to the status of cancer cell lines. We developed a novel strategy to build a GRN-based classifier, called a GRN-classifier. The proposed GRN-classifier estimates GRNs and classifies cell lines simultaneously, where the gene network is estimated to minimize error in gene network estimation and the negative log-likelihood for classifying cell lines. Thus, we can identify biological status-specific gene regulatory systems, enabling us to achieve biologically reliable interpretation of the classification. We also propose an algorithm to implement the GRN-classifier based on coordinate descent update. Monte Carlo simulations were conducted to examine performance of the GRN-classifier. Results: Our strategy provides effective results in feature selection in the classification model and edge selection in gene network estimation. The GRN-classifier also shows outstanding classification accuracy. We apply the GRN-classifier to classify cancer cell lines into anticancer drug-related status, that is, 5-fluorouracil (5-FU)-sensitive/resistant and 5-FU target/nontarget cancer cell lines. We then identified 5-FU markers based on 5-FU-related status classification-specific gene networks. The mechanisms of the identified markers were verified through literature survey. Our results suggest that the molecular interplay between MYOF and AHNAK2 may play a crucial role in drug resistance and can provide information on the chemotherapy efficiency of 5-FU. It is also suggested that suppression of the identified 5-FU markers, including MYOF/AHNAK2 and AKR1C1/AKR1C3 may improve 5-FU resistance of cancer cell lines.

RN-Autoencoder: Reduced Noise Autoencoder for classifying imbalanced cancer genomic data

BACKGROUND

In the current genomic era, gene expression datasets have become one of the main tools utilized in cancer classification. Both curse of dimensionality and class imbalance problems are inherent characteristics of these datasets. These characteristics have a negative impact on the performance of most classifiers when used to classify cancer using genomic datasets.

RESULTS

This paper introduces Reduced Noise-Autoencoder (RN-Autoencoder) for pre-processing imbalanced genomic datasets for precise cancer classification. Firstly, RN-Autoencoder solves the curse of dimensionality problem by utilizing the autoencoder for feature reduction and hence generating new extracted data with lower dimensionality. In the next stage, RN-Autoencoder introduces the extracted data to the well-known Reduced Noise-Synthesis Minority Over Sampling Technique (RN- SMOTE) that efficiently solve the problem of class imbalance in the extracted data. RN-Autoencoder has been evaluated using different classifiers and various imbalanced datasets with different imbalance ratios. The results proved that the performance of the classifiers has been improved with RN-Autoencoder and outperformed the performance with original data and extracted data with percentages based on the classifier, dataset and evaluation metric. Also, the performance of RN-Autoencoder has been compared to the performance of the current state of the art and resulted in an increase up to 18.017, 19.183, 18.58 and 8.87% in terms of test accuracy using colon, leukemia, Diffuse Large B-Cell Lymphoma (DLBCL) and Wisconsin Diagnostic Breast Cancer (WDBC) datasets respectively.

CONCLUSION

RN-Autoencoder is a model for cancer classification using imbalanced gene expression datasets. It utilizes the autoencoder to reduce the high dimensionality of the gene expression datasets and then handles the class imbalance using RN-SMOTE. RN-Autoencoder has been evaluated using many different classifiers and many different imbalanced datasets. The performance of many classifiers has improved and some have succeeded in classifying cancer with 100% performance in terms of all used metrics. In addition, RN-Autoencoder outperformed many recent works using the same datasets.

Cancer Segmentation by Entropic Analysis of Ordered Gene Expression Profiles

The availability of massive gene expression data has been challenging in terms of how to cure, process, and extract useful information. Here, we describe the use of entropic measures as discriminating criteria in cancer using the whole data set of gene expression levels. These methods were applied in classifying samples between tumor and normal type for 13 types of tumors with a high success ratio. Using gene expression, ordered by pathways, results in complexity-entropy diagrams. The map allows the clustering of the tumor and normal types samples, with a high success rate for nine of the thirteen, studied cancer types. Further analysis using information distance also shows good discriminating behavior, but, more importantly, allows for discriminating between cancer types. Together, our results allow the classification of tissues without the need to identify relevant genes or impose a particular cancer model. The used procedure can be extended to classification problems beyond the reported results.

Computational Tactics for Precision Cancer Network Biology

Network biology has garnered tremendous attention in understanding complex systems of cancer, because the mechanisms underlying cancer involve the perturbations in the specific function of molecular networks, rather than a disorder of a single gene. In this article, we review the various computational tactics for gene regulatory network analysis, focused especially on personalized anti-cancer therapy. This paper covers three major topics: (1) cell line's (or patient's) cancer characteristics specific gene regulatory network estimation, which enables us to reveal molecular interplays under varying conditions of cancer characteristics of cell lines (or patient); (2) computational approaches to interpret the multitudinous and massive networks; (3) network-based application to uncover molecular mechanisms of cancer and related marker identification. We expect that this review will help readers understand personalized computational network biology that plays a significant role in precision cancer medicine.

Machine Learning-Based Peripheral Artery Disease Identification Using Laboratory-Based Gait Data

Peripheral artery disease (PAD) manifests from atherosclerosis, which limits blood flow to the legs and causes changes in muscle structure and function, and in gait performance. PAD is underdiagnosed, which delays treatment and worsens clinical outcomes. To overcome this challenge, the purpose of this study is to develop machine learning (ML) models that distinguish individuals with and without PAD. This is the first step to using ML to identify those with PAD risk early. We built ML models based on previously acquired overground walking biomechanics data from patients with PAD and healthy controls. Gait signatures were characterized using ankle, knee, and hip joint angles, torques, and powers, as well as ground reaction forces (GRF). ML was able to classify those with and without PAD using Neural Networks or Random Forest algorithms with 89% accuracy (0.64 Matthew's Correlation Coefficient) using all laboratory-based gait variables. Moreover, models using only GRF variables provided up to 87% accuracy (0.64 Matthew's Correlation Coefficient). These results indicate that ML models can classify those with and without PAD using gait signatures with acceptable performance. Results also show that an ML gait signature model that uses GRF features delivers the most informative data for PAD classification.

Deep learning techniques for cancer classification using microarray gene expression data

Cancer is one of the top causes of death globally. Recently, microarray gene expression data has been used to aid in cancer’s effective and early detection. The use of DNA microarray technology to uncover information from the expression levels of thousands of genes has enormous promise. The DNA microarray technique can determine the levels of thousands of genes simultaneously in a single experiment. The analysis of gene expression is critical in many disciplines of biological study to obtain the necessary information. This study analyses all the research studies focused on optimizing gene selection for cancer detection using artificial intelligence. One of the most challenging issues is figuring out how to extract meaningful information from massive databases. Deep Learning architectures have performed efficiently in numerous sectors and are used to diagnose many other chronic diseases and to assist physicians in making medical decisions. In this study, we have evaluated the results of different optimizers on a RNA sequence dataset. The Deep learning algorithm proposed in the study classifies five different forms of cancer, including kidney renal clear cell carcinoma (KIRC), Breast Invasive Carcinoma (BRCA), lung adenocarcinoma (LUAD), Prostate Adenocarcinoma (PRAD) and Colon Adenocarcinoma (COAD). The performance of different optimizers like Stochastic gradient descent (SGD), Root Mean Squared Propagation (RMSProp), Adaptive Gradient Optimizer (AdaGrad), and Adaptive Momentum (AdaM). The experimental results gathered on the dataset affirm that AdaGrad and Adam. Also, the performance analysis has been done using different learning rates and decay rates. This study discusses current advancements in deep learning-based gene expression data analysis using optimized feature selection methods.

Artificial Neural Network Assisted Cancer Risk Prediction of Oral Precancerous Lesions

The incidence of oral cancer is still increasing. It has become very common in patients with malignant tumors, which has forced medical personnel to continuously explore its treatment methods. What kind of method can effectively and correctly diagnose the disease in the early stage and improve the survival rate has become one of the research topics that have attracted much attention. Aiming at this problem, it has great research significance for the field of oral precancerous lesions diagnosis. With the in-depth research on oral precancerous diagnosis, the research on artificial neural network (ANN) in medical diagnosis is gradually carried out. Its performance advantage is of great significance to solve the problem of early and correct disease diagnosis. This paper aimed to investigate the application of ANN-assisted cancer risk prediction method in risk prediction of oral precancerous lesions. Through the analysis and research of ANN and oral cancer, the construction of oral cancer risk prediction model was applied to solve the problem of improving the survival rate of oral cancer patients. In this paper, ANN and oral precancerous lesions were analyzed, the performance of the algorithm was experimentally analyzed, and the relevant theoretical formulas were used to explain. The results showed that the method had higher accuracy than traditional forecasting methods. When $N=2$ , the output accuracy was above 90%. It can be seen that the algorithm can meet the needs of the diagnosis of high-risk groups of oral cancer lesions, and the diagnosis efficiency and patient survival rate has been greatly improved.

Construction of a neural network diagnostic model and investigation of immune infiltration characteristics for Crohn’s disease

Objective: Crohn’s disease (CD), a chronic recurrent illness, is a type of inflammatory bowel disease whose incidence and prevalence rates are gradually increasing. However, there is no universally accepted criterion for CD diagnosis. The aim of this study was to create a diagnostic prediction model for CD and identify immune cell infiltration features in CD.

Methods: In this study, gene expression microarray datasets were obtained from the Gene Expression Omnibus (GEO) database. Then, we identified differentially expressed genes (DEGs) between 178 CD and 38 control cases. Enrichment analysis of DEGs was also performed to explore the biological role of DEGs. Moreover, the “randomForest” package was applied to select core genes that were used to create a neural network model. Finally, in the training cohort, we used CIBERSORT to evaluate the immune landscape between the CD and normal groups.

Results: The results of enrichment analysis revealed that these DEGs may be involved in biological processes associated with immunity and inflammatory responses. Moreover, the top 3 hub genes in the protein-protein interaction network were IL-1β, CCL2, and CXCR2. The diagnostic model allowed significant discrimination with an area under the ROC curve of 0.984 [95% confidence interval: 0.971–0.993]. A validation cohort (GSE36807) was utilized to ensure the reliability and applicability of the model. In addition, the immune infiltration analysis indicated nine different immune cell types were significantly different between the CD and healthy control groups.

Conclusion: In summary, this study offers a novel insight into the diagnosis of CD and provides potential biomarkers for the precise treatment of CD.

Fucose as a potential therapeutic molecule against the immune-mediated inflammation in IgA nepharopathy: An unrevealed link

Background

IgA nephropathy (IgAN) is an autoimmune disease that affects people of any age and is an important cause of end-stage renal disease. However, the pathogenesis and pathophysiology of IgAN is not clear. This article aimed to explore the immune-mediated inflammation and genetic mechanisms in IgAN.

Methods

The transcriptome sequencing data of IgAN glomeruli in the Gene Expression Omnibus database were downloaded. Single-sample gene set enrichment analysis was used to estimate the immune microenvironment of the merged microarray data and GSE141295. IgAN samples were divided into two clusters by cluster analysis. “limma” and “DEseq2” package in R were used to identify differentially expressed genes (DEGs). The weighted gene co-expression network analysis (WGCNA) was used to identify the co-expression modules related to inflammation in IgAN. R software package “clusterProfiler” was used for enrichment analysis, whereas Short Time-Series Expression Miner (STEM) analysis was used to identify the trend of gene expression. Machine-learn (ML) was performed using the shiny app. Finally, Drug Signatures Database (DSigDB) was used to identify potential molecules for treating IgAN.

Results

The infiltration of macrophages in IgAN glomeruli was increased, whereas CD4+ T cells, especially inducedregulatory T cells (iTregs) were decreased. A total of 1,104 common DEGs were identified from the merged data and GSE141295. Brown module was identified to have the highest inflammatory correlation with IgAN using WGCNA, and 15 hub genes were screened from this module. Among these 15 hub genes, 14 increased with the severity of IgAN inflammation based on STEM analysis. Neural network (nnet) is considered as the best model to predict the severity of IgAN. Fucose identified from DSigDB has a potential biological activity to treat IgAN.

Conclusion

The increase of macrophages and the decrease of iTregs in glomeruli represent the immune-mediated inflammation of IgAN, and fucose may be a potential therapeutic molecule against IgAN because it affects genes involved in the severe inflammation of IgAN.

PredictiveNetwork: predictive gene network estimation with application to gastric cancer drug response-predictive network analysis

Background

Gene regulatory networks have garnered a large amount of attention to understand disease mechanisms caused by complex molecular network interactions. These networks have been applied to predict specific clinical characteristics, e.g., cancer, pathogenicity, and anti-cancer drug sensitivity. However, in most previous studies using network-based prediction, the gene networks were estimated first, and predicted clinical characteristics based on pre-estimated networks. Thus, the estimated networks cannot describe clinical characteristic-specific gene regulatory systems. Furthermore, existing computational methods were developed from algorithmic and mathematics viewpoints, without considering network biology.

Results

To effectively predict clinical characteristics and estimate gene networks that provide critical insights into understanding the biological mechanisms involved in a clinical characteristic, we propose a novel strategy for predictive gene network estimation. The proposed strategy simultaneously performs gene network estimation and prediction of the clinical characteristic. In this strategy, the gene network is estimated with minimal network estimation and prediction errors. We incorporate network biology by assuming that neighboring genes in a network have similar biological functions, while hub genes play key roles in biological processes. Thus, the proposed method provides interpretable prediction results and enables us to uncover biologically reliable marker identification. Monte Carlo simulations shows the effectiveness of our method for feature selection in gene estimation and prediction with excellent prediction accuracy. We applied the proposed strategy to construct gastric cancer drug-responsive networks.

Conclusion

We identified gastric drug response predictive markers and drug sensitivity/resistance-specific markers, AKR1B10, AKR1C3, ANXA10, and ZNF165, based on GDSC data analysis. Our results for identifying drug sensitive and resistant specific molecular interplay are strongly supported by previous studies. We expect that the proposed strategy will be a useful tool for uncovering crucial molecular interactions involved a specific biological mechanism, such as cancer progression or acquired drug resistance.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12859-022-04871-z.

MetastaSite: Predicting metastasis to different sites using deep learning with gene expression data

Deep learning has massive potential in predicting phenotype from different omics profiles. However, deep neural networks are viewed as black boxes, providing predictions without explanation. Therefore, the requirements for these models to become interpretable are increasing, especially in the medical field. Here we propose a computational framework that takes the gene expression profile of any primary cancer sample and predicts whether patients' samples are primary (localized) or metastasized to the brain, bone, lung, or liver based on deep learning architecture. Specifically, we first constructed an AutoEncoder framework to learn the non-linear relationship between genes, and then DeepLIFT was applied to calculate genes' importance scores. Next, to mine the top essential genes that can distinguish the primary and metastasized tumors, we iteratively added ten top-ranked genes based upon their importance score to train a DNN model. Then we trained a final multi-class DNN that uses the output from the previous part as an input and predicts whether samples are primary or metastasized to the brain, bone, lung, or liver. The prediction performances ranged from AUC of 0.93-0.82. We further designed the model's workflow to provide a second functionality beyond metastasis site prediction, i.e., to identify the biological functions that the DL model uses to perform the prediction. To our knowledge, this is the first multi-class DNN model developed for the generic prediction of metastasis to various sites.

Early detection of Alzheimer's disease using single nucleotide polymorphisms analysis based on gradient boosting tree

Alzheimer's disease (AD) is a degenerative disorder that attacks nerve cells in the brain. AD leads to memory loss and cognitive & intellectual impairments that can influence social activities and decision-making. The most common type of human genetic variation is single nucleotide polymorphisms (SNPs). SNPs are beneficial markers of complex gene-disease. Many common and serious diseases, such as AD, have associated SNPs. Detection of SNP biomarkers linked with AD could help in the early prediction and diagnosis of this disease. The main objective of this paper is to predict and diagnose AD based on SNPs biomarkers with high classification accuracy in the early stages. One of the most concerning problems is the high number of features. Thus, the paper proposes a comprehensive framework for early AD detection and detecting the most significant genes based on SNPs analysis. Usage of machine learning (ML) techniques to identify new biomarkers of AD is also suggested. In the proposed system, two feature selection techniques are separately checked: the information gain filter and Boruta wrapper. The two feature selection techniques were used to select the most significant genes related to AD in this system. Filter methods measure the relevance of features by their correlation with dependent variables, while wrapper methods measure the usefulness of a subset of features by training a model on it. Gradient boosting tree (GBT) has been applied on all AD genetic data of neuroimaging initiative phase 1 (ADNI-1) and Whole-Genome Sequencing (WGS) datasets by using two feature selection techniques. In the whole-genome approach ADNI-1, results revealed that the GBT learning algorithm scored an overall accuracy of 99.06% in the case of using Boruta feature selection. Using information gain feature selection, the proposed system achieved an average accuracy of 94.87%. The results show that the proposed system is preferable for the early detection of AD. Also, the results revealed that the Boruta wrapper feature selection is superior to the information gain filter technique.

Identification of Secondary Breast Cancer in Vital Organs through the Integration of Machine Learning and Microarrays

Breast cancer includes genetic and environmental factors and is the most prevalent malignancy in women contributing to the pathogenesis and progression of cancer. Breast cancer prognosis metastasizes towards bones, the liver, brain, and lungs, and is the main cause of death in patients. Furthermore, the selection of features and classification is significant in microarray data analysis, which suffers from huge time consumption. To address these issues, this research uniquely integrates machine learning and microarrays to identify secondary breast cancer in vital organs. This work firstly imputes the missing values using K-nearest neighbors and improves the recursive feature elimination with cross-validation (RFECV) using the random forest method. Secondly, the class imbalance is handled by employing K-means synthetic object oversampling technique (SMOTE) to balance minority class and prevent noise. We successfully identified the 16 most essential Entrez gene ids responsible for predicting metastatic locations in the bones, brain, liver, and lungs. Extensive experiments are conducted on NCBI Gene Expression Omnibus GSE14020 and GSE54323 datasets. The proposed methods have handled class imbalance, prevented noise, and appropriately reduced time consumption. Reliable results were obtained on four classification models: decision tree; K-nearest neighbors; random forest; and support vector machine. Results are presented having considered confusion matrices, accuracy, ROC-AUC and PR-AUC, and F1-score.

Xprediction: Explainable EGFR-TKIs response prediction based on drug sensitivity specific gene networks

In recent years, drug sensitivity prediction has garnered a great deal of attention due to the growing interest in precision medicine. Several computational methods have been developed for drug sensitivity prediction and the identification of related markers. However, most previous studies have ignored genetic interaction, although complex diseases (e.g., cancer) involve many genes intricately connected in a molecular network rather than the abnormality of a single gene. To effectively predict drug sensitivity and understand its mechanism, we propose a novel strategy for explainable drug sensitivity prediction based on sample-specific gene regulatory networks, designated Xprediction. Our strategy first estimates sample-specific gene regulatory networks that enable us to identify the molecular interplay underlying varying clinical characteristics of cell lines. We then, predict drug sensitivity based on the estimated sample-specific gene regulatory networks. The predictive models are based on machine learning approaches, i.e., random forest, kernel support vector machine, and deep neural network. Although the machine learning models provide remarkable results for prediction and classification, we cannot understand how the models reach their decisions. In other words, the methods suffer from the black box problem and thus, we cannot identify crucial molecular interactions that involve drug sensitivity-related mechanisms. To address this issue, we propose a method that describes the importance of each molecular interaction for the drug sensitivity prediction result. The proposed method enables us to identify crucial gene-gene interactions and thereby, interpret the prediction results based on the identified markers. To evaluate our strategy, we applied Xprediction to EGFR-TKIs prediction based on drug sensitivity specific gene regulatory networks and identified important molecular interactions for EGFR-TKIs prediction. Our strategy effectively performed drug sensitivity prediction compared with prediction based on the expression levels of genes. We also verified through literature, the EGFR-TKIs-related mechanisms of a majority of the identified markers. We expect our strategy to be a useful tool for predicting tasks and uncovering complex mechanisms related to pharmacological profiles, such as mechanisms of acquired drug resistance or sensitivity of cancer cells.

Red Fox Optimizer with Data-Science-Enabled Microarray Gene Expression Classification Model

Microarray data examination is a relatively new technology that intends to determine the proper treatment for various diseases and a precise medical diagnosis by analyzing a massive number of genes in various experimental conditions. The conventional data classification techniques suffer from overfitting and the high dimensionality of gene expression data. Therefore, the feature (gene) selection approach plays a vital role in handling a high dimensionality of data. Data science concepts can be widely employed in several data classification problems, and they identify different class labels. In this aspect, we developed a novel red fox optimizer with deep-learning-enabled microarray gene expression classification (RFODL-MGEC) model. The presented RFODL-MGEC model aims to improve classification performance by selecting appropriate features. The RFODL-MGEC model uses a novel red fox optimizer (RFO)-based feature selection approach for deriving an optimal subset of features. Moreover, the RFODL-MGEC model involves a bidirectional cascaded deep neural network (BCDNN) for data classification. The parameters involved in the BCDNN technique were tuned using the chaos game optimization (CGO) algorithm. Comprehensive experiments on benchmark datasets indicated that the RFODL-MGEC model accomplished superior results for subtype classifications. Therefore, the RFODL-MGEC model was found to be effective for the identification of various classes for high-dimensional and small-scale microarray data.

Feature selection methods on gene expression microarray data for cancer classification: A systematic review

This systematic review provides researchers interested in feature selection (FS) for processing microarray data with comprehensive information about the main research directions for gene expression classification conducted during the recent seven years. A set of 132 researches published by three different publishers is reviewed. The studied papers are categorized into nine directions based on their objectives. The FS directions that received various levels of attention were then summarized. The review revealed that 'propose hybrid FS methods' represented the most interesting research direction with a percentage of 34.9%, while the other directions have lower percentages that ranged from 13.6% down to 3%. This guides researchers to select the most competitive research direction. Papers in each category are thoroughly reviewed based on six perspectives, mainly: method(s), classifier(s), dataset(s), dataset dimension(s) range, performance metric(s), and result(s) achieved.

Biomarker Identification From Gene Expression Based on Symmetrical Uncertainty

In this paper, we present an effective information theoretic feature selection method, Symmetrical Uncertainty to classify gene expression microarray data and detect biomarkers from it. Here, Information Gain and Symmetrical Uncertainty contribute for ranking the features. Based on computed values of Symmetrical Uncertainty, features were sorted from most informative to least informative ones. Then, the top features from the sorted list are passed to Random Forest, Logistic Regression and other well-known classifiers with Leave-One-Out cross validation to construct the best classification model(s) and accordingly select the most important genes from microarray datasets. Obtained results in terms of classification accuracy, running time, root mean square error and other parameters computed on Leukemia and Colon cancer datasets demonstrate the effectiveness of the proposed approach. The proposed method is relatively much faster than many other wrapper or ensemble methods.

Genetic variations analysis for complex brain disease diagnosis using machine learning techniques: opportunities and hurdles

BACKGROUND AND OBJECTIVES

This paper presents an in-depth review of the state-of-the-art genetic variations analysis to discover complex genes associated with the brain's genetic disorders. We first introduce the genetic analysis of complex brain diseases, genetic variation, and DNA microarrays. Then, the review focuses on available machine learning methods used for complex brain disease classification. Therein, we discuss the various datasets, preprocessing, feature selection and extraction, and classification strategies. In particular, we concentrate on studying single nucleotide polymorphisms (SNP) that support the highest resolution for genomic fingerprinting for tracking disease genes. Subsequently, the study provides an overview of the applications for some specific diseases, including autism spectrum disorder, brain cancer, and Alzheimer's disease (AD). The study argues that despite the significant recent developments in the analysis and treatment of genetic disorders, there are considerable challenges to elucidate causative mutations, especially from the viewpoint of implementing genetic analysis in clinical practice. The review finally provides a critical discussion on the applicability of genetic variations analysis for complex brain disease identification highlighting the future challenges.

METHODS

We used a methodology for literature surveys to obtain data from academic databases. Criteria were defined for inclusion and exclusion. The selection of articles was followed by three stages. In addition, the principal methods for machine learning to classify the disease were presented in each stage in more detail.

RESULTS

It was revealed that machine learning based on SNP was widely utilized to solve problems of genetic variation for complex diseases related to genes.

CONCLUSIONS

Despite significant developments in genetic diseases in the past two decades of the diagnosis and treatment, there is still a large percentage in which the causative mutation cannot be determined, and a final genetic diagnosis remains elusive. So, we need to detect the variations of the genes related to brain disorders in the early disease stages.

Machine Learning Protocols in Early Cancer Detection Based on Liquid Biopsy: A Survey

With the advances of liquid biopsy technology, there is increasing evidence that body fluid such as blood, urine, and saliva could harbor the potential biomarkers associated with tumor origin. Traditional correlation analysis methods are no longer sufficient to capture the high-resolution complex relationships between biomarkers and cancer subtype heterogeneity. To address the challenge, researchers proposed machine learning techniques with liquid biopsy data to explore the essence of tumor origin together. In this survey, we review the machine learning protocols and provide corresponding code demos for the approaches mentioned. We discuss algorithmic principles and frameworks extensively developed to reveal cancer mechanisms and consider the future prospects in biomarker exploration and cancer diagnostics.

Underwater localization system based on visible-light communications using neural networks

Underwater localization using visible-light communications is proposed based on neural networks (NNs) estimation of received signal strength (RSS). Our proposed work compromises two steps: data collection and NN training. First, data are collected with the aid of Zemax OpticStudio Monte Carlo ray tracing software, where we configure 40,000 receivers in a $100\;{\rm m} \times 100\;{\rm m}$ area in order to measure the channel gain for each detector in seawater. The channel gains represent the input data set to the NN, while the output of the NN is the coordinates of each detector based on the RSS intensity technique. Next, an NN system is built and trained with the aid of Orange data mining software. Several trials for NN implementations are performed, and the best training algorithms, activation functions, and number of neurons are determined. In addition, several performance measures are considered in order to evaluate the robustness of the proposed network. Specifically, we evaluate the following parameters: classification accuracy (CA), area under the curve (AUC), training time, testing time, F1, precision, recall, logloss, and specificity. The corresponding measures are as follows: 99.1% for AUC and 98.7% for CA, F1, precision, and recall. Further, the performance results of logloss and specificity are 7.3% and 99.3% respectively.

Machine Learning Methods for Identifying Atrial Fibrillation Cases and Their Predictors in Patients With Hypertrophic Cardiomyopathy: The HCM-AF-Risk Model

Background

Hypertrophic cardiomyopathy (HCM) patients have a high incidence of atrial fibrillation (AF) and increased stroke risk, even with low CHA₂DS₂-VASc (congestive heart failure, hypertension, age diabetes, previous stroke/transient ischemic attack) scores. Hence, there is a need to understand the pathophysiology of AF/stroke in HCM. In this retrospective study, we develop and apply a data-driven, machine learning–based method to identify AF cases, and clinical/imaging features associated with AF, using electronic health record data.

Methods

HCM patients with documented paroxysmal/persistent/permanent AF (n = 191) were considered AF cases, and the remaining patients in sinus rhythm (n = 640) were tagged as No-AF. We evaluated 93 clinical variables; the most informative variables useful for distinguishing AF from No-AF cases were selected based on the 2-sample t test and the information gain criterion.

Results

We identified 18 highly informative variables that are positively (n = 11) and negatively (n = 7) correlated with AF in HCM. Next, patient records were represented via these 18 variables. Data imbalance resulting from the relatively low number of AF cases was addressed via a combination of oversampling and undersampling strategies. We trained and tested multiple classifiers under this sampling approach, showing effective classification. Specifically, an ensemble of logistic regression and naïve Bayes classifiers, trained based on the 18 variables and corrected for data imbalance, proved most effective for separating AF from No-AF cases (sensitivity = 0.74, specificity = 0.70, C-index = 0.80).

Conclusions

Our model (HCM-AF-Risk Model) is the first machine learning–based method for identification of AF cases in HCM. This model demonstrates good performance, addresses data imbalance, and suggests that AF is associated with a more severe cardiac HCM phenotype.

Real-time neural network based predictor for cov19 virus spread

Since the epidemic outbreak in early months of 2020 the spread of COVID-19 has grown rapidly in most countries and regions across the World. Because of that, SARS-CoV-2 was declared as a Public Health Emergency of International Concern (PHEIC) on January 30, 2020, by The World Health Organization (WHO). That’s why many scientists are working on new methods to reduce further growth of new cases and, by intelligent patients allocation, reduce number of patients per doctor, what can lead to more successful treatments. However to properly manage the COVID-19 spread there is a need for real-time prediction models which can reliably support various decisions both at national and international level. The problem in developing such system is the lack of general knowledge how the virus spreads and what would be the number of cases each day. Therefore prediction model must be able to conclude the situation from past data in the way that results will show a future trend and will possibly closely relate to the real numbers. In our opinion Artificial Intelligence gives a possibility to do it. In this article we present a model which can work as a part of an online system as a real-time predictor to help in estimation of COVID-19 spread. This prediction model is developed using Artificial Neural Networks (ANN) to estimate the future situation by the use of geo-location and numerical data from past 2 weeks. The results of our model are confirmed by comparing them with real data and, during our research the model was correctly predicting the trend and very closely matching the numbers of new cases in each day.

Biological interpretation of deep neural network for phenotype prediction based on gene expression

BACKGROUND

The use of predictive gene signatures to assist clinical decision is becoming more and more important. Deep learning has a huge potential in the prediction of phenotype from gene expression profiles. However, neural networks are viewed as black boxes, where accurate predictions are provided without any explanation. The requirements for these models to become interpretable are increasing, especially in the medical field.

RESULTS

We focus on explaining the predictions of a deep neural network model built from gene expression data. The most important neurons and genes influencing the predictions are identified and linked to biological knowledge. Our experiments on cancer prediction show that: (1) deep learning approach outperforms classical machine learning methods on large training sets; (2) our approach produces interpretations more coherent with biology than the state-of-the-art based approaches; (3) we can provide a comprehensive explanation of the predictions for biologists and physicians.

CONCLUSION

We propose an original approach for biological interpretation of deep learning models for phenotype prediction from gene expression data. Since the model can find relationships between the phenotype and gene expression, we may assume that there is a link between the identified genes and the phenotype. The interpretation can, therefore, lead to new biological hypotheses to be investigated by biologists.

Autoencoded DNA methylation data to predict breast cancer recurrence: Machine learning models and gene-weight significance

Breast cancer is the most frequent cancer in women and the second most frequent overall after lung cancer. Although the 5-year survival rate of breast cancer is relatively high, recurrence is also common which often involves metastasis with its consequent threat for patients. DNA methylation-derived databases have become an interesting primary source for supervised knowledge extraction regarding breast cancer. Unfortunately, the study of DNA methylation involves the processing of hundreds of thousands of features for every patient. DNA methylation is featured by High Dimension Low Sample Size which has shown well-known issues regarding feature selection and generation. Autoencoders (AEs) appear as a specific technique for conducting nonlinear feature fusion. Our main objective in this work is to design a procedure to summarize DNA methylation by taking advantage of AEs. Our proposal is able to generate new features from the values of CpG sites of patients with and without recurrence. Then, a limited set of relevant genes to characterize breast cancer recurrence is proposed by the application of survival analysis and a pondered ranking of genes according to the distribution of their CpG sites. To test our proposal we have selected a dataset from The Cancer Genome Atlas data portal and an AE with a single-hidden layer. The literature and enrichment analysis (based on genomic context and functional annotation) conducted regarding the genes obtained with our experiment confirmed that all of these genes were related to breast cancer recurrence.

Handling imbalanced medical image data: A deep-learning-based one-class classification approach

In clinical settings, a lot of medical image datasets suffer from the imbalance problem which hampers the detection of outliers (rare health care events), as most classification methods assume an equal occurrence of classes. In this way, identifying outliers in imbalanced datasets has become a crucial issue. To help address this challenge, one-class classification, which focuses on learning a model using samples from only a single given class, has attracted increasing attention. Previous one-class modeling usually uses feature mapping or feature fitting to enforce the feature learning process. However, these methods are limited for medical images which usually have complex features. In this paper, a novel method is proposed to enable deep learning models to optimally learn single-class-relevant inherent imaging features by leveraging the concept of imaging complexity. We investigate and compare the effects of simple but effective perturbing operations applied to images to capture imaging complexity and to enhance feature learning. Extensive experiments are performed on four clinical datasets to show that the proposed method outperforms four state-of-the-art methods.

Complex MIMO RBF Neural Networks for Transmitter Beamforming over Nonlinear Channels

The use of beamforming for efficient transmission has already been successfully implemented in practical systems and is absolutely necessary to even further increase spectral and energy efficiencies in some configurations of the next-generation wireless systems and for low earth orbit satellites. A remarkable capacity increase is then achieved and spectral congestion is minimized. In this context, this article proposes a novel complex multiple-input multiple-output radial basis function neural network (CMM-RBF) for transmitter beamforming, based on the phase transmittance radial basis function neural network (PTRBFNN). The proposed CMM-RBF is compared with the least mean square (LMS) algorithm for beamforming with six dipoles arranged in a uniform and circular array and with 16 dipoles in a 2D-grid array. Simulation results show that the proposed solution presents lower steady-state mean squared error, faster convergence rate and enhanced half-power beamwidth (HPBW) when compared with the LMS algorithm in a nonlinear scenario.