Individuals with coronavirus disease 2019 (COVID-19) infection present in a variety of ways, ranging from asymptomatic or mild cough, to organ failure or death. One of the major challenges for the medical community is the quick and accurate determination of how COVID-19 will progress in an individual. Herein, we introduce a new Cut-and-Solve based feature selection program for identifying predictive feature sets in heterogeneous data. We analyze proteomics data from Washington University to identify models ranging in size from a single feature up to five. Validation of logistic regression models using area under the curve (AUC) were applied for both a holdout data set and an independent data set from Massachusetts General Hospital. A variety of known and novel biomarkers for COVID-19 severity were identified. The best model for predicting severe (ventilation or death) vs. non-severe infection is achieved for CALCOCO2 and STC1, with an average AUC=0.81. Based on the known severity markers, several different proteomic pathways are identified. Enrichment analysis indicates activity associated with inflammatory response, as well as myelination and cardiac function.

MOTIVATION: Establishment of a statistical association between microbiome features and clinical outcomes is of growing interest because of the potential for yielding insights into biological mechanisms and pathogenesis. Extracting microbiome features that are relevant for a disease is challenging and existing variable selection methods are limited due to large number of risk factor variables from microbiome sequence data and their complex biological structure. RESULTS: We propose a tree-based scanning method, Selection of Models for the Analysis of Risk factor Trees (referred to as SMART-scan), for identifying taxonomic groups that are associated with a disease or trait. SMART-scan is a model selection technique that uses a predefined taxonomy to organize the large pool of possible predictors into optimized groups, and hierarchically searches and determines variable groups for association test. We investigate the statistical properties of SMART-scan through simulations, in comparison to a regular single-variable analysis and three commonly-used variable selection methods, stepwise regression, least absolute shrinkage and selection operator (LASSO) and classification and regression tree (CART). When there are taxonomic group effects in the data, SMART-scan can significantly increase power by using bacterial taxonomic information to split large numbers of variables into groups. Through an application to microbiome data from a vervet monkey diet experiment, we demonstrate that SMART-scan can identify important phenotype-associated taxonomic features missed by single-variable analysis, stepwise regression, LASSO and CART.

Hundreds of genetic markers have shown associations with various complex diseases, yet the "missing heritability" remains alarmingly elusive. Combinatorial interactions may account for a substantial portion of this missing heritability, but their discoveries have been impeded by computational complexity and genetic heterogeneity. We present BlocBuster, a novel systems-level approach that efficiently constructs genome-wide, allele-specific networks that accurately segregate homogenous combinations of genetic factors, tests the associations of these combinations with the given phenotype, and rigorously validates the results using a series of unbiased validation methods. BlocBuster employs a correlation measure that is customized for single nucleotide polymorphisms and returns a multi-faceted collection of values that captures genetic heterogeneity. We applied BlocBuster to analyze psoriasis, discovering a combinatorial pattern with an odds ratio of 3.64 and Bonferroni-corrected p-value of 5.01×10(-16). This pattern was replicated in independent data, reflecting robustness of the method. In addition to improving prediction of disease susceptibility and broadening our understanding of the pathogenesis underlying psoriasis, these results demonstrate BlocBuster's potential for discovering combinatorial genetic associations within heterogeneous genome-wide data, thereby transcending the limiting "small effects" produced by individual markers examined in isolation.

MOTIVATION: Inference of haplotypes from genotype data is crucial and challenging for many vitally important studies. The first, and most critical step, is the ascertainment of a biologically sound model to be optimized. Many models that have been proposed rely partially or entirely on reducing the number of unique haplotypes in the solution. RESULTS: This article examines the parsimony of haplotypes using known haplotypes as well as genotypes from the HapMap project. Our study reveals that there are relatively few unique haplotypes, but not always the least possible, for the datasets with known solutions. Furthermore, we show that there are frequently very large numbers of parsimonious solutions, and the number increases exponentially with increasing cardinality. Moreover, these solutions are quite varied, most of which are not consistent with the true solutions. These results quantify the limitations of the Pure Parsimony model and demonstrate the imperative need to consider additional properties for haplotype inference models. At a higher level, and with broad applicability, this article illustrates the power of combinatorial methods to tease out imperfections in a given biological model.

SUMMARY: We present the software implementation of the tree scanning method to detect associations between genetic haplotypes and quantitative traits, utilizing the evolutionary history of the haplotypes, in samples of unrelated individuals. AVAILABILITY: The program is available free of charge, under the GNU General Public License. A package including C source code, a Makefile, and Windows (DOS) and Macintosh binaries, can be downloaded from http://darwin.uvigo.es