T-Recs: Stable Selection of Dynamically Formed Groups of Features With Application to Prediction of Clinical Outcomes

Abstract

Feature selection is used extensively in biomedical research for biomarker identification and patient classification, both of which are essential steps in developing personalized medicine strategies. However, the structured nature of the biological datasets and high correlation of variables frequently yield multiple equally optimal signatures, thus making traditional feature selection methods unstable. Features selected based on one cohort of patients, may not work as well in another cohort. In addition, biologically important features may be missed due to selection of other co-clustered features We propose a new method, Tree-guided Recursive Cluster Selection (T-ReCS), for efficient selection of grouped features. T-ReCS significantly improves predictive stability while maintains the same level of accuracy. T-ReCS does not require an a priori knowledge of the clusters like group-lasso and also can handle "orphan" features (not belonging to a cluster). T-ReCS can be used with categorical or survival target variables. Tested on simulated and real expression data from breast cancer and lung diseases and survival data, T-ReCS selected stable cluster features without significant loss in classification accuracy.

Respiratory Microbiome Profiling for Etiologic Diagnosis of Pneumonia in Mechanically Ventilated Patients

“…These findings extended our univariate taxonomic analyses in that they identified the taxa that are directly linked to positive cultures (not simply correlated) and highlighted sequenced bacterial taxa as the strongest explanatory variables of culture positivity. To examine the ability of 16S taxonomic data alone to predict culture positivity, we used the Markov blanket around the culture-positivity variable as a feature selection method ( Huang et al, 2015 ). The taxonomy-based classifier yielded mean accuracy of 82.3% ( SD = 7%) ( Table 3 ), indicating proof-of-concept utility for use of sequencing in clinical practice for predicting culture results, if sequencing results were available real-time.…”

Learning mixed graphical models with separate sparsity parameters and stability-based model selection

“…Traditional univariate approaches for feature selection exist as well, but they also often operate on a single data type. In addition, due to the high dimensionality and co-linearity of biological data, markers selected by these standard feature selection algorithms can be unstable and lack biological relevance [ 2 ], a problem that has recently been addressed directly [ 3 ]. Many existing models that do integrate different data types make heavy use of prior knowledge [ 4 , 5 ] and as such are not easily extendable to clinical and other data that are not well studied.…”

Redes bayesianas con algoritmos basados en restricciones, scores e híbridos aplicados al problema de clasificación

“…En general, se busca identificar un subconjunto de variables predictoras que son relevantes con respecto a una tarea específica; por ejemplo, en regresión y clasificación se busca seleccionar y retener el subconjunto de variables predictoras con el más alto poder predictivo. Se han desarrollado algoritmos que generan múltiples conjuntos de variables equivalentes (Huang et al, 2014). El algoritmo Statistically Equivalent Signature, SES, (Tsamardinos et al, 2013) permite identificar múltiples subconjuntos de variables con rendimientos estadísticamente equivalentes.…”