An efficiency analysis of high-order combinations of gene-gene interactions using multifactor-dimensionality reduction

From Machine Learning to Patient Outcomes: A Comprehensive Review of AI in Pancreatic Cancer

Pancreatic cancer is a highly aggressive and difficult-to-detect cancer with a poor prognosis. Late diagnosis is common due to a lack of early symptoms, specific markers, and the challenging location of the pancreas. Imaging technologies have improved diagnosis, but there is still room for improvement in standardizing guidelines. Biopsies and histopathological analysis are challenging due to tumor heterogeneity. Artificial Intelligence (AI) revolutionizes healthcare by improving diagnosis, treatment, and patient care. AI algorithms can analyze medical images with precision, aiding in early disease detection. AI also plays a role in personalized medicine by analyzing patient data to tailor treatment plans. It streamlines administrative tasks, such as medical coding and documentation, and provides patient assistance through AI chatbots. However, challenges include data privacy, security, and ethical considerations. This review article focuses on the potential of AI in transforming pancreatic cancer care, offering improved diagnostics, personalized treatments, and operational efficiency, leading to better patient outcomes.

Dimensionality reduction approach for many-objective epistasis analysis

In epistasis analysis, single-nucleotide polymorphism-single-nucleotide polymorphism interactions (SSIs) among genes may, alongside other environmental factors, influence the risk of multifactorial diseases. To identify SSI between cases and controls (i.e. binary traits), the score for model quality is affected by different objective functions (i.e. measurements) because of potential disease model preferences and disease complexities. Our previous study proposed a multiobjective approach-based multifactor dimensionality reduction (MOMDR), with the results indicating that two objective functions could enhance SSI identification with weak marginal effects. However, SSI identification using MOMDR remains a challenge because the optimal measure combination of objective functions has yet to be investigated. This study extended MOMDR to the many-objective version (i.e. many-objective MDR, MaODR) by integrating various disease probability measures based on a two-way contingency table to improve the identification of SSI between cases and controls. We introduced an objective function selection approach to determine the optimal measure combination in MaODR among 10 well-known measures. In total, 6 disease models with and 40 disease models without marginal effects were used to evaluate the general algorithms, namely those based on multifactor dimensionality reduction, MOMDR and MaODR. Our results revealed that the MaODR-based three objective function model, correct classification rate, likelihood ratio and normalized mutual information (MaODR-CLN) exhibited the higher 6.47% detection success rates (Accuracy) than MOMDR and higher 17.23% detection success rates than MDR through the application of an objective function selection approach. In a Wellcome Trust Case Control Consortium, MaODR-CLN successfully identified the significant SSIs (P < 0.001) associated with coronary artery disease. We performed a systematic analysis to identify the optimal measure combination in MaODR among 10 objective functions. Our combination detected SSIs-based binary traits with weak marginal effects and thus reduced spurious variables in the score model. MOAI is freely available at https://sites.google.com/view/maodr/home.

MOAI: a multi-outcome interaction identification approach reveals an interaction between vaspin and carcinoembryonic antigen on colorectal cancer prognosis

Identifying and characterizing the interaction between risk factors for multiple outcomes (multi-outcome interaction) has been one of the greatest challenges faced by complex multifactorial diseases. However, the existing approaches have several limitations in identifying the multi-outcome interaction. To address this issue, we proposed a multi-outcome interaction identification approach called MOAI. MOAI was motivated by the limitations of estimating the interaction simultaneously occurring in multi-outcomes and by the success of Pareto set filter operator for identifying multi-outcome interaction. MOAI permits the identification for the interaction of multiple outcomes and is applicable in population-based study designs. Our experimental results exhibited that the existing approaches are not effectively used to identify the multi-outcome interaction, whereas MOAI obviously exhibited superior performance in identifying multi-outcome interaction. We applied MOAI to identify the interaction between risk factors for colorectal cancer (CRC) in both metastases and mortality prognostic outcomes. An interaction between vaspin and carcinoembryonic antigen (CEA) was found, and the interaction indicated that patients with CRC characterized by higher vaspin (≥30%) and CEA (≥5) levels could simultaneously increase both metastases and mortality risk. The immunostaining evidence revealed that determined multi-outcome interaction could effectively identify the difference between non-metastases/survived and metastases/deceased patients, which offers multi-prognostic outcome risk estimation for CRC. To our knowledge, this is the first report of a multi-outcome interaction associated with a complex multifactorial disease. MOAI is freely available at https://sites.google.com/view/moaitool/home.

An improved fuzzy set-based multifactor dimensionality reduction for detecting epistasis

OBJECTIVE

Epistasis identification is critical for determining susceptibility to human genetic diseases. The rapid development of technology has enabled scalability to make multifactor dimensionality reduction (MDR) measurements an effective calculation tool that achieves superior detection. However, the classification of high-risk (H) or low-risk (L) groups in multidrug resistance operations calls for extensive research.

METHODS AND MATERIAL

In this study, an improved fuzzy sigmoid (FS) method using the membership degree in MDR (FSMDR) was proposed for solving the limitations of binary classification. The FS method combined with MDR measurements yielded an improved ability to distinguish similar frequencies of potential multifactor genotypes.

RESULTS

We compared our results with other MDR-based methods and FSMDR achieved superior detection rates on simulated data sets. The results indicated that the fuzzy classifications can provide insight into the uncertainty of H/L classification in MDR operation.

CONCLUSION

FSMDR successfully detected significant epistasis of coronary artery disease in the Wellcome Trust Case Control Consortium data set.

Class Balanced Multifactor Dimensionality Reduction to Detect Gene-Gene Interactions

Detecting gene-gene interactions in single-nucleotide polymorphism data is vital for understanding disease susceptibility. However, existing approaches may be limited by the sample size in case-control studies. Herein, we propose a balance approach for the multifactor dimensionality reduction (BMDR) method to increase the accuracy of estimates of the prediction error rate in small samples. BMDR explicitly selects the best model by evaluating the average of prediction error rates over k-fold cross-validation without cross-validation consistency selection. In this study, we used several epistatic models with and without marginal effects under different parameter settings (heritability and minor allele frequencies) to evaluate the performance of existing approaches. Using simulated data sets, BMDR successfully detected gene-gene interactions, particularly for data sets with small sample sizes. A large data set was obtained from the Wellcome Trust Case Control Consortium, and results indicated that BMDR could effectively detect significant gene-gene interactions.

Multiobjective multifactor dimensionality reduction to detect SNP-SNP interactions

Motivation

Single-nucleotide polymorphism (SNP)-SNP interactions (SSIs) are popular markers for understanding disease susceptibility. Multifactor dimensionality reduction (MDR) can successfully detect considerable SSIs. Currently, MDR-based methods mainly adopt a single-objective function (a single measure based on contingency tables) to detect SSIs. However, generally, a single-measure function might not yield favorable results due to potential model preferences and disease complexities.

Approach

This study proposes a multiobjective MDR (MOMDR) method that is based on a contingency table of MDR as an objective function. MOMDR considers the incorporated measures, including correct classification and likelihood rates, to detect SSIs and adopts set theory to predict the most favorable SSIs with cross-validation consistency. MOMDR enables simultaneously using multiple measures to determine potential SSIs.

Results

Three simulation studies were conducted to compare the detection success rates of MOMDR and single-objective MDR (SOMDR), revealing that MOMDR had higher detection success rates than SOMDR. Furthermore, the Wellcome Trust Case Control Consortium dataset was analyzed by MOMDR to detect SSIs associated with coronary artery disease. Availability and implementation: MOMDR is freely available at https://goo.gl/M8dpDg.

Supplementary information

Supplementary data are available at Bioinformatics online.

PBMDR: A particle swarm optimization-based multifactor dimensionality reduction for the detection of multilocus interactions

Studies on multilocus interactions have mainly investigated the associations between genetic variations from the related genes and histopathological tumor characteristics in patients. However, currently, the identification and characterization of susceptibility genes for complex diseases remain a great challenge for geneticists. In this study, a particle swarm optimization (PSO)-based multifactor dimensionality reduction (MDR) approach was proposed, denoted by PBMDR. MDR was used to detect multilocus interactions based on the PSO algorithm. A test data set was simulated from the genotype frequencies of 26 SNPs from eight breast-cancer-related gene. In simulated disease models, we demonstrated that PBMDR outperforms existing global optimization algorithms in terms of its ability to explore and power to detect specific SNP-genotype combinations. In addition, the PBMDR algorithm was compared with other algorithms, including PSO and chaotic PSOs, and the results revealed that the PBMDR algorithm yielded higher accuracy and chi-square values than other algorithms did.

Catfish Taguchi-Based Binary Differential Evolution Algorithm for Analyzing Single Nucleotide Polymorphism Interactions in Chronic Dialysis

Single-nucleotide polymorphism (SNP)-SNP interactions are crucial for understanding the association between disease-related multifactorials for disease analysis. Existing statistical methods for determining such interactions are limited by the considerable computation required for evaluating all potential associations between disease-related multifactorials. Identifying SNP-SNP interactions is thus a major challenge in genetic association studies. This paper proposes a catfish Taguchi-based binary differential evolution (CT-BDE) algorithm for identifying SNP-SNP interactions. In the search space, the catfish effect prevents the premature convergence of the population, and the Taguchi method improves the search ability of the BDE algorithm. Hence, the proposed algorithm enables obtaining a favorable solution regarding the identification of high-order SNP-SNP interactions. Additionally, the proposed algorithm applies an effective fitness function derived from a multifactor dimensionality reduction (MDR) operation to evaluate the solutions from BDE-based algorithms. Simulated and real data sets were used to evaluate the ability of several BDE-based algorithms in identifying specific SNP-SNP interactions. We compared the fitness function derived from the MDR operation with that derived according to the difference between cases and controls, by using the different BDE-based algorithms. The results showed that the proposed CT-BDE algorithm applying the fitness function derived from the MDR operation exhibited a superior ability in identifying SNP-SNP interactions compared with the other BDE-based algorithms.

Multiple-Criteria Decision Analysis-Based Multifactor Dimensionality Reduction for Detecting Gene-Gene Interactions

Gene-gene interactions (GGIs) are important markers for determining susceptibility to a disease. Multifactor dimensionality reduction (MDR) is a popular algorithm for detecting GGIs and primarily adopts the correct classification rate (CCR) to assess the quality of a GGI. However, CCR measurement alone may not successfully detect certain GGIs because of potential model preferences and disease complexities. In this study, multiple-criteria decision analysis (MCDA) based on MDR was named MCDA-MDR and proposed for detecting GGIs. MCDA facilitates MDR to simultaneously adopt multiple measures within the two-way contingency table of MDR to assess GGIs; the CCR and rule utility measure were employed. Cross-validation consistency was adopted to determine the most favorable GGIs among the Pareto sets. Simulation studies were conducted to compare the detection success rates of the MDR-only-based measure and MCDA-MDR, revealing that MCDA-MDR had superior detection success rates. The Wellcome Trust Case Control Consortium dataset was analyzed using MCDA-MDR to detect GGIs associated with coronary artery disease, and MCDA-MDR successfully detected numerous significant GGIs (p < 0.001). MCDA-MDR performance assessment revealed that the applied MCDA successfully enhanced the GGI detection success rate of the MDR-based method compared with MDR alone.

Collective feature selection to identify crucial epistatic variants

Background

Machine learning methods have gained popularity and practicality in identifying linear and non-linear effects of variants associated with complex disease/traits. Detection of epistatic interactions still remains a challenge due to the large number of features and relatively small sample size as input, thus leading to the so-called "short fat data" problem. The efficiency of machine learning methods can be increased by limiting the number of input features. Thus, it is very important to perform variable selection before searching for epistasis. Many methods have been evaluated and proposed to perform feature selection, but no single method works best in all scenarios. We demonstrate this by conducting two separate simulation analyses to evaluate the proposed collective feature selection approach.

Results

Through our simulation study we propose a collective feature selection approach to select features that are in the "union" of the best performing methods. We explored various parametric, non-parametric, and data mining approaches to perform feature selection. We choose our top performing methods to select the union of the resulting variables based on a user-defined percentage of variants selected from each method to take to downstream analysis. Our simulation analysis shows that non-parametric data mining approaches, such as MDR, may work best under one simulation criteria for the high effect size (penetrance) datasets, while non-parametric methods designed for feature selection, such as Ranger and Gradient boosting, work best under other simulation criteria. Thus, using a collective approach proves to be more beneficial for selecting variables with epistatic effects also in low effect size datasets and different genetic architectures. Following this, we applied our proposed collective feature selection approach to select the top 1% of variables to identify potential interacting variables associated with Body Mass Index (BMI) in ~ 44,000 samples obtained from Geisinger's MyCode Community Health Initiative (on behalf of DiscovEHR collaboration).

Conclusions

In this study, we were able to show that selecting variables using a collective feature selection approach could help in selecting true positive epistatic variables more frequently than applying any single method for feature selection via simulation studies. We were able to demonstrate the effectiveness of collective feature selection along with a comparison of many methods in our simulation analysis. We also applied our method to identify non-linear networks associated with obesity.

Synergistic Association of Genetic Variants with Environmental Risk Factors in Susceptibility to Essential Hypertension

AIMS

Essential hypertension (EH) is a disease in which both environment and genes have an important role. This study was designed to identify the interaction model between genetic variants and environmental risk factors that most highly potentiates EH development.

METHODS

We performed a case-control study with 1641 participants (mean age 50.6 ± 8.1 years), specifically 848 patients with EH and 793 controls, adjusted for gender and age. Traditional risk factors, biochemical and genetic parameters, including the genotypic discrimination of 14 genetic variants previously associated with EH, were investigated. Multifactorial dimensionality reduction (MDR) software was used to analyze gene-environment interactions. Validation was performed using logistic regression analysis with environmental risk factors, significant genetic variants, and the best MDR model.

RESULTS

The best model indicates that the interactions among the ADD1 rs4961 640T allele, diabetes, and obesity (body mass index ≥30) increase approximately four-fold the risk of EH (odds ratio = 3.725; 95% confidence interval: 2.945-4.711; p < 0.0001).

CONCLUSION

This work showed that the interaction between the ADD1 rs4961 variant, obesity, and the presence of diabetes increased the susceptibility to EH four-fold. In these circumstances, lifestyle adjustment and diabetes control should be intensified in patients who carry the ADD1 variant.

A Multiple Interaction Analysis Reveals ADRB3 as a Potential Candidate for Gallbladder Cancer Predisposition via a Complex Interaction with Other Candidate Gene Variations

Gallbladder cancer is the most common and a highly aggressive biliary tract malignancy with a dismal outcome. The pathogenesis of the disease is multifactorial, comprising the combined effect of multiple genetic variations of mild consequence along with numerous dietary and environmental risk factors. Previously, we demonstrated the association of several candidate gene variations with GBC risk. In this study, we aimed to identify the combination of gene variants and their possible interactions contributing towards genetic susceptibility of GBC. Here, we performed Multifactor-Dimensionality Reduction (MDR) and Classification and Regression Tree Analysis (CRT) to investigate the gene–gene interactions and the combined effect of 14 SNPs in nine genes (DR4 (rs20576, rs6557634); FAS (rs2234767); FASL (rs763110); DCC (rs2229080, rs4078288, rs7504990, rs714); PSCA (rs2294008, rs2978974); ADRA2A (rs1801253); ADRB1 (rs1800544); ADRB3 (rs4994); CYP17 (rs2486758)) involved in various signaling pathways. Genotyping was accomplished by PCR-RFLP or Taqman allelic discrimination assays. SPSS software version 16.0 and MDR software version 2.0 were used for all the statistical analysis. Single locus investigation demonstrated significant association of DR4 (rs20576, rs6557634), DCC (rs714, rs2229080, rs4078288) and ADRB3 (rs4994) polymorphisms with GBC risk. MDR analysis revealed ADRB3 (rs4994) to be crucial candidate in GBC susceptibility that may act either alone (p < 0.0001, CVC = 10/10) or in combination with DCC (rs714 and rs2229080, p < 0.0001, CVC = 9/10). Our CRT results are in agreement with the above findings. Further, in-silico results of studied SNPs advocated their role in splicing, transcriptional and/or protein coding regulation. Overall, our result suggested complex interactions amongst the studied SNPs and ADRB3 rs4994 as candidate influencing GBC susceptibility.