Unsupervised Algorithms for Microarray Sample Stratification

Fratello, Michele; Cattelani, Luca; Federico, Antonio; Pavel, Alisa; Scala, Giovanni; Serra, Angela; Greco, Dario

doi:10.1007/978-1-0716-1839-4_9

Cited by 3 publications

(4 citation statements)

References 126 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There is a wide range of statistical methods [2] that have had a great impact on the field of Computational Statistics. These include cluster analysis and classification techniques [3,4] that can be used to separate groups with similar characteristics. Correspondence Analysis (CA) [5] separates groups without considering affinities but is an excellent support for global visualization in clinical cases.…”

Section: Introductionmentioning

confidence: 99%

Group Classification for the Search and Identification of Related Patterns Using a Variety of Multivariate Techniques

Boukichou-Abdelkader,

Montero-Alonso,

Muñoz-García

2024

Computation

View full text Add to dashboard Cite

Recently, many methods and algorithms have been developed that can be quickly adapted to different situations within a population of interest, especially in the health sector. Success has been achieved by generating better models and higher-quality results to facilitate decision making, as well as to propose new diagnostic procedures and treatments adapted to each patient. These models can also improve people’s quality of life, dissuade bad health habits, reinforce good habits, and modify the pre-existing ones. In this sense, the objective of this study was to apply supervised and unsupervised classification techniques, where the clustering algorithm was the key factor for grouping. This led to the development of three optimal groups of clinical pattern based on their characteristics. The supervised classification methods used in this study were Correspondence (CA) and Decision Trees (DT), which served as visual aids to identify the possible groups. At the same time, they were used as exploratory mechanisms to confirm the results for the existing information, which enhanced the value of the final results. In conclusion, this multi-technique approach was found to be a feasible method that can be used in different situations when there are sufficient data. It was thus necessary to reduce the dimensional space, provide missing values for high-quality information, and apply classification models to search for patterns in the clinical profiles, with a view to grouping the patients efficiently and accurately so that the clinical results can be applied in other research studies.

show abstract

Section: Introductionmentioning

confidence: 99%

Group Classification for the Search and Identification of Related Patterns Using a Variety of Multivariate Techniques

Boukichou-Abdelkader,

Montero-Alonso,

Muñoz-García

2024

Computation

View full text Add to dashboard Cite

show abstract

“…In addition, the comparison of molecular alteration profiles allows to identify similarities between phenotypic entities and to make conclusions about possible phenotypic changes of an exposure ( Kinaret et al 2020b ). Transcriptomics data are complex and prone to technical and biological variability and noise ( Raser and O’Shea 2005 , Freytag et al 2015 , Federico et al 2020 , Fratello et al 2022 ). Therefore many variables need to be considered when comparing expression profiles, especially coming from different datasets or (biological) systems.…”

Section: Introductionmentioning

confidence: 99%

“…Therefore the individual gene view is replaced by a “similar gene” view, where instead of considering genes individually, a set of genes are grouped together based on multi-level prior knowledge. This gene grouping is used to create a feature vector for each experimental instance, which can be used in downstream analysis, such as clustering or machine learning (ML) applications, where often a numeric feature vector is needed as input ( Serra et al 2020 , Fratello et al 2022 ). This is in contrast to many functional enrichment applications, where individual pathway names are returned, that cannot be directly provided as input to such downstream ML applications.…”

Section: Introductionmentioning

confidence: 99%

KNeMAP: a network mapping approach for knowledge-driven comparison of transcriptomic profiles

et al. 2023

Self Cite

View full text Add to dashboard Cite

Motivation Transcriptomic data can be used to describe the mechanism of action (MOA) of a chemical compound. However, omics data tend to be complex and prone to noise, making the comparison of different datasets challenging. Often, transcriptomic profiles are compared at the level of individual gene expression values, or sets of differentially expressed genes. Such approaches can suffer from underlying technical and biological variance, such as the biological system exposed on or the machine/method used to measure gene expression data, technical errors and further neglect the relationships between the genes. We propose a network mapping approach for knowledge-driven comparison of transcriptomic profiles (KNeMAP), which combines genes into similarity groups based on multiple levels of prior information, hence adding a higher level view onto the individual gene view. When comparing KNeMAP with fold change (expression) based and deregulated gene set based methods, KNeMAP was able to group compounds with higher accuracy with respect to prior information as well as is less prone to noise corrupted data. Result We applied KNeMAP to analyze the Connectivity Map dataset, where the gene expression changes of three cell lines were analyzed after treatment with 676 drugs as well as the Fortino et al. dataset where two cell lines with 31 nanomaterials were analyzed. Although the expression profiles across the biological systems are highly different, KNeMAP was able to identify sets of compounds that induce similar molecular responses when exposed on the same biological system. Availability Relevant data and the KNeMAP function is available at: https://github.com/fhaive/KNeMAP and 10.5281/zenodo.7334711. Supplementary information Supplementary data are available at Bioinformatics online.

show abstract

“…To help researchers with analyzing and understanding the considerable amount of available microarray data before it becomes obsolete, several bio-statistical [ 15 , 16 , 17 , 18 , 19 , 20 , 21 ], machine learning [ 22 , 23 , 24 , 25 , 26 , 27 , 28 ], and statistical [ 29 , 30 , 31 , 32 ] methods are used to interpret the results’ biological meaning better. Unsupervised learning [ 33 , 34 , 35 , 36 , 37 ] techniques have been widely applied in the analysis of microarray studies to reveal hidden patterns in the data [ 38 , 39 , 40 , 41 , 42 ]. In unsupervised learning, only the input examples are provided to the model, without any expected output, to make it learn some hidden structure in the input data independently.…”

Section: Introductionmentioning

confidence: 99%

A Python Clustering Analysis Protocol of Genes Expression Data Sets

Agapito

Milano

Cannataro

2022

Genes

View full text Add to dashboard Cite

Gene expression and SNPs data hold great potential for a new understanding of disease prognosis, drug sensitivity, and toxicity evaluations. Cluster analysis is used to analyze data that do not contain any specific subgroups. The goal is to use the data itself to recognize meaningful and informative subgroups. In addition, cluster investigation helps data reduction purposes, exposes hidden patterns, and generates hypotheses regarding the relationship between genes and phenotypes. Cluster analysis could also be used to identify bio-markers and yield computational predictive models. The methods used to analyze microarrays data can profoundly influence the interpretation of the results. Therefore, a basic understanding of these computational tools is necessary for optimal experimental design and meaningful data analysis. This manuscript provides an analysis protocol to effectively analyze gene expression data sets through the K-means and DBSCAN algorithms. The general protocol enables analyzing omics data to identify subsets of features with low redundancy and high robustness, speeding up the identification of new bio-markers through pathway enrichment analysis. In addition, to demonstrate the effectiveness of our clustering analysis protocol, we analyze a real data set from the GEO database. Finally, the manuscript provides some best practice and tips to overcome some issues in the analysis of omics data sets through unsupervised learning.

show abstract

Unsupervised Algorithms for Microarray Sample Stratification

Cited by 3 publications

References 126 publications

Group Classification for the Search and Identification of Related Patterns Using a Variety of Multivariate Techniques

Group Classification for the Search and Identification of Related Patterns Using a Variety of Multivariate Techniques

KNeMAP: a network mapping approach for knowledge-driven comparison of transcriptomic profiles

A Python Clustering Analysis Protocol of Genes Expression Data Sets

Contact Info

Product

Resources

About