Heterogeneity impacts biomarker discovery for precision medicine

Abstract

Precision medicine is advancing patient care for complex human diseases. Discovery of biomarkers to diagnose specific subtypes within a heterogeneous diseased population is a key step towards realizing the benefits of precision medicine. However, popular statistical methods for evaluating candidate biomarkers, fold change and AUC, were designed for homogeneous data and we evaluate their performance here. In general, these metrics overlook nearly ‘ideal’ biomarkers when they represent less than half of the diseased population. We introduce a new metric to address this shortfall and run a series of trials comprised of simulated and biological data.

A machine-learning evaluation of biomarkers designed for the future of precision medicine

“…[10][11][12][13] It is patently clear that AUC and FC fail to identify biomarkers for subtypes of heterogeneous diseases when the subtype comprises less than half of the entire group of diseased cases. 1 Subtypes have inherently low true positive rates, which sabotage AUC assessments, and are lost in summary statistics such as FC. Instead of focusing on these traditional measurements, we have developed tools based on the assumption that a subtype will form a secondary cluster within the data values for the diseased cases.…”

“…Our recent approach, BCD, is based upon the statistical characteristics of the data and shows great improvements for large sample sizes. 1 This manuscript introduces a machine-learning approach, DBD, based on clustering that is suitable for more moderate sample sizes. In summary, DBD provides a unique and effective method for screening real-valued data to identify biomarkers associated with subtypes of heterogeneous diseases.…”

“…We previously demonstrated the inability for FC and AUC to capture signals for subtypes when they comprise less than 50% of the diseased individuals. 1 FC is unable to capture the subtypes as only the mean or median is used in the computation and subtype signals are lost. AUC's dependence upon the true positive rate cripples its evaluation as the upper limit on this rate is the percentage of individuals in the subset.…”

“…Cognizant of the limitations of the true positive rate, as well as the related measurements in the confusion matrix, for detecting subtypes we previously proposed a bimodality coefficient difference (BCD) evaluation. 1 Based on the assumption that a subtype will create a secondary peak in the distribution of the data for diseased cases, BCD computes the bimodality coefficient for each group of diseased cases and normal controls and is set to the difference of these two values. The bimodality coefficient is based on statistical characteristics of the data, including skewness, cardinality, and kurtosis.…”

Microbiota Implications in Endocrine-Related Diseases: From Development to Novel Therapeutic Approaches

“…Additionally, exploring specific genetic markers within the microbial genome adds another layer of complexity to endocrine-related diagnostics, enabling a more targeted approach and potentially paving the way for personalized interventions [297]. This integrative use of diverse biomarkers enhances the precision and comprehensiveness of microbiota-endocrine assessments, fostering a deeper understanding of this intricate interplay for both diagnostic and therapeutic purposes [298][299][300].…”