Datasets in which measurements of two (or more) types are obtained from a common set of samples arise in many scientific applications. A common problem in the exploratory analysis of such data is to identify groups of features of different data types that are strongly associated. A bimodule is a pair (A,B) of feature sets from two data types such that the aggregate cross-correlation between the features in A and those in B is large. A bimodule (A,B) is stable if A coincides with the set of features that have significant aggregate correlation with the features in B, and vice-versa. This paper proposes an iterative-testing based bimodule search procedure (BSP) to identify stable bimodules. Compared to existing methods for detecting cross-correlated features, BSP was the best at recovering true bimodules with sufficient signal, while limiting the false discoveries. In addition, we applied BSP to the problem of expression quantitative trait loci (eQTL) analysis using data from the GTEx consortium. BSP identified several thousand SNP-gene bimodules. While many of the individual SNP-gene pairs appearing in the discovered bimodules were identified by standard eQTL methods, the discovered bimodules revealed genomic subnetworks that appeared to be biologically meaningful and worthy of further scientific investigation.

The18F-AV-1451 PET tracer binds to tau, an Alzheimer's disease biomarker. The standardized uptake value ratio (SUVR) 80-100 min window is widely used to quantify tau binding, although18F-AV-1451 continues increasing relative to a reference region in regions with tau deposition. Left uncorrected, acquisition time inaccuracies can lead to errors from -4% to 6% in 20-min SUVR measurements in subjects with Alzheimer's disease. In 40 subjects with scans from 75-115 min following18F-AV-1451 injection, we created 20-min reconstructions (4×5 min) of start-times ranging from 75-85 min, as proxies of offset scans and calculated the mean in regions of interest (ROIs). We developed a segmented least squares (SLS) method to obtain error-minimizing weighting coefficients for18F-AV-1451 ROIs that best predict SUVR 80-100 from weighted means of SUVRs from offset start-times. We compared residual errors of our SLS method to those in: 1) uncorrected offset 20-min-SUVRs; 2) the mean of five-min frames within the 80-100 window; and 3) a least-squares interpolation method. We evaluated errors induced by start-time offset on SUVRs for each method. TheSLS, which corrected using least-squares coefficients of 5-min components, consistently reduced errors across all offset starttimes. Effect size analysis for simulated clinical longitudinal18F-AV-1451 drug trials showed that uncorrected 20-min offset SUVRs would require up to 20% more participants to detect treatment effects compared with using SLS. Correction of SUVR scan-time errors by SLS minimizes errors compared with other correction methods and may be extended to other scanners and tracers.