Numero: a statistical framework to define multivariable subgroups in complex population-based datasets

Abstract

Large-scale epidemiological and population data provide opportunities to identify subgroups of people who are at risk of disease or exposed to adverse environments. Clustering algorithms are popular data-driven tools to identify these subgroups; however, relying exclusively on algorithms may not produce the best results if the dataset does not have a clustered structure. For this reason, we propose a framework (the R-library Numero) that combines the self-organizing map algorithm, permutation analysis for statistical evidence and a final expert-driven subgrouping step. We used Numero to define subgroups in two examples without an obvious clustering structure: a biomedical dataset of kidney disease and another dataset of community-level socioeconomic indicators. We benchmarked the Numero subgroupings against popular clustering algorithms (principal components, K-means and hierarchical clustering). The Numero subgroupings were more intuitive and easier to interpret without losing mathematical quality. Therefore, we expect Numero to be useful for exploratory analyses of population-based epidemiological datasets.

Commentary: Data-driven subgrouping in epidemiology and medicine

“…Groop et al 3 and suggested some correspondence between the independently specified data-driven clusters even though the data used were fundamentally different, i.e., genetic loci 2 versus clinical and biomarker data at the time of diabetes diagnosis. 3 The examples of SOM analyses demonstrated by Mäkinen and co-workers, 5,6,8 as well as the recent other clustering applications in diabetes; 2,3 suggest alluring potential for data-driven subgroup analyses in epidemiology and medicine. These types of approaches are likely to inform on the fundamental genetic and metabolic variation defining the complexity of polygenic diseases.…”

“…In this Issue Mäkinen and co-workers present a Software Application Profile for an opensource R library, titled Numero, which would be a versatile and powerful tool for the abovementioned types of subgrouping needs. 8 Numero provides a three-step framework that combines the self-organizing map (SOM) algorithm, permutation analyses for statistical evidence and an expert-driven final subgrouping decision. 5,8 Numero can handle both continuous and categorical variables in situations where there is no intrinsic clustering in the data and it creates data-driven statistically validated two-dimensional visualisations without explicit boundaries for the subgroups.…”

“…8 Numero provides a three-step framework that combines the self-organizing map (SOM) algorithm, permutation analyses for statistical evidence and an expert-driven final subgrouping decision. 5,8 Numero can handle both continuous and categorical variables in situations where there is no intrinsic clustering in the data and it creates data-driven statistically validated two-dimensional visualisations without explicit boundaries for the subgroups. This adds unavoidable complexity to the use of the SOM algorithm for subgrouping and calls for substance experts in each application for balanced interpretations.…”

Perspectivas do uso de mineração de dados e aprendizado de máquina em saúde e segurança no trabalho

“…No estudo de Pearce et al 21 , o SOM foi utilizado para analisar a qualidade de ar com base em séries históricas. O trabalho realizado por Gao et al 22 utiliza uma variação do SOM para realizar análises exploratórias nas quais o algoritmo cria subgrupos de populações com características similares, usando como base de dados os registros médicos de doentes renais. O algoritmo gera um mapa bidimensional representando cinco dimensões escolhidas e exibe a intensidade de concentração das variáveis selecionadas, assim como as probabilidades de mortalidade em oito anos, de acordo com os perfis identificados pelo SOM.…”

“…Gao et al 22 Olson et al 18 Lee et al 19 Quadro Com os resultados obtidos na pesquisa realizada na Biblioteca Virtual em Saúde e nos demais estudos citados neste ensaio, elaborou-se uma relação de finalidades, aplicações e estudos exemplificativos (Quadro 2).…”