Search

Integrating multi-omics datasets is critical for microbiome research, but multiple statistical challenges can confound traditional correlation techniques. We solve this problem by using neural networks to estimate the conditional probability that each molecule is present given the presence of each specific microbe. We show with known environmental (desert biological soil crust wetting) and clinical (cystic fibrosis lung) examples, our ability to recover microbe-metabolite relationships, and demonstrate how the method can discover relationships between microbially-produced metabolites and inflammatory bowel disease.

Recently, Quinn and Erb et al [1] made the case that when used correctly, correlation and proportionality can outperform MMvec when identifying microbe-metabolite interactions. We revisit this comparison and show that the proposed correlation and proportionality are outperformed by MMvec on real data due to their inability to deal with sparsity commonly observed in microbiome and metabolome datasets., bioRxiv

CRLF2 overexpression in B-ALL patients with an IGH-CRLF2 translocation activates JAK-STAT, PI3K and ERK/MAPK signaling pathways. Although inhibitors of these pathways are available, investigating alternate targets could reduce treatment-associated toxicities. Comparing RNA-seq from IGH-CRLF2 and non-translocated patients we defined a translocation gene signature. Next, we assembled a B-ALL cancer-specific regulatory network using 529 B-ALL patient samples from the NCI TARGET database coupled with priors generated from ATAC-seq peak TF-motif analysis. The network was used to infer differential changes in TF activities predicted to control IGH-CRLF2 deregulated genes, thereby enabling identification of translocation-associated pathways and potential new therapeutic targets.

The Taos trough is one of several tectonically active cratonic basins related to late Paleozoic ARM deformation. Two competing basin models for the Taos trough were tested in order to better discern the evolution of the ancestral Rocky Mountains in northern New Mexico. The opposing models were pure flexural foreland basin versus multiple basins separated by intrabasinal uplifts. The composite section from Highway 518 Holman grade and Forest Service Road 723, in northern New Mexico, records 949 m of middle Pennsylvanian fluvial deltaic deposits. Conodonts recovered from the base and top of the composite section indicate that the deposition occurred during the early to middle Desmoinesian. Sedimentologic and stratigraphic data indicate that the succession is cyclic with three separate cycle types recorded: 1) A-type, 2) B-type, and 3) C-type. Each cycle is defined by cycle-top facies composed of variable water depths. Twenty one cycles are exposed in the study area with 11 being A-type, 7 B-type, and 3 C-Type. Control on the cycles was largely variations in tectonic subsidence while correlation of relative water depth records to far-field cyclic successions in the Donets Basin suggests that eustasy largely modulated the cyclic successions. Sediment transport data was used to discern the likely provenance region for the sediment in the study area. The data indicate a predominantly westerly transport direction, supporting an eastern source. Sediment dispersal indicators in fluvial and deltaic facies are easiest to reconcile with a nearby highland, rather than the Sierra Grande uplift (~ 80-100 km away). Thus, the El Oro-Rincon uplift is the preferred eastern highland shedding sediment into the study area. Previous studies, on the west side of the Taos trough axis, record an easterly sediment transport direction and were previously used to bolster the inference of a flexural model. Our westerly sourced data combined with the easterly sourced suggests that clastic material funneled into a central axis of the Taos trough. A backstripped subsidence model indicates that basin subsidence began during the Early Pennsylvanian, and that rates increased rapidly during the Early Desmoinesian. Subsidence rates greatly decreased in the Late Pennsylvanian, indicating a period of tectonic quiescence, but uptick again in the Early Permian. The activity created nearly 5km of total Late Paleozoic subsidence in the eastern Taos trough. The tectonic subsidence curve compares to both strike-slip and flexural derived foreland basin curves and indicates that the basin likely experienced a combination of both strike-slip and thrust systems. Comparison of the subsidence model and sediment transport data of the Taos trough indicates that the Taos trough is difficult to characterize as a purely flexural basin or a strike-slip basin. A potential unifying structural model may involve transfer of Early Pennsylvanian sinistral slip from the Pecos-Picuris fault into the basal thrust of the intrabasinal El Oro-Rincon uplift. In this scenario the bend in the Pecos-Picuris fault system south of the ancestral Brazos uplift allows for sinistral slip to transfer to the basin center resulting in a series of basement involved thrusts within the basin.

For investigations into fate specification and cell rearrangements in live images of preimplantation embryos, automated and accurate 3D instance segmentation of nuclei is invaluable; however, the performance of segmentation methods is limited by the images' low signal-to-noise ratio and high voxel anisotropy and the nuclei's dense packing and variable shapes. Supervised machine learning approaches have the potential to radically improve segmentation accuracy but are hampered by a lack of fully annotated 3D data. In this work, we first establish a novel mouse line expressing near-infrared nuclear reporter H2B-miRFP720. H2B-miRFP720 is the longest wavelength nuclear reporter in mice and can be imaged simultaneously with other reporters with minimal overlap. We then generate a dataset, which we call BlastoSPIM, of 3D microscopy images of H2B-miRFP720-expressing embryos with ground truth for nuclear instance segmentation. Using BlastoSPIM, we benchmark the performance of five convolutional neural networks and identify Stardist-3D as the most accurate instance segmentation method across preimplantation development. Stardist-3D, trained on BlastoSPIM, performs robustly up to the end of preimplantation development (> 100 nuclei) and enables studies of fate patterning in the late blastocyst. We, then, demonstrate BlastoSPIM's usefulness as pre-train data for related problems. BlastoSPIM and its corresponding Stardist-3D models are available at: blastospim.flatironinstitute.org.