Dataset Information

Species-level deconvolution of metagenome assemblies with Hi-C-based contact probability maps.

ABSTRACT: Microbial communities consist of mixed populations of organisms, including unknown species in unknown abundances. These communities are often studied through metagenomic shotgun sequencing, but standard library construction methods remove long-range contiguity information; thus, shotgun sequencing and de novo assembly of a metagenome typically yield a collection of contigs that cannot readily be grouped by species. Methods for generating chromatin-level contact probability maps, e.g., as generated by the Hi-C method, provide a signal of contiguity that is completely intracellular and contains both intrachromosomal and interchromosomal information. Here, we demonstrate how this signal can be exploited to reconstruct the individual genomes of microbial species present within a mixed sample. We apply this approach to two synthetic metagenome samples, successfully clustering the genome content of fungal, bacterial, and archaeal species with more than 99% agreement with published reference genomes. We also show that the Hi-C signal can secondarily be used to create scaffolded genome assemblies of individual eukaryotic species present within the microbial community, with higher levels of contiguity than some of the species' published reference genomes.

SUBMITTER: Burton JN

PROVIDER: S-EPMC4455782 | biostudies-literature | 2014 May

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Species-level deconvolution of metagenome assemblies with Hi-C-based contact probability maps.

Burton Joshua N JN Liachko Ivan I Dunham Maitreya J MJ Shendure Jay J

G3 (Bethesda, Md.) 20140522 7

Microbial communities consist of mixed populations of organisms, including unknown species in unknown abundances. These communities are often studied through metagenomic shotgun sequencing, but standard library construction methods remove long-range contiguity information; thus, shotgun sequencing and de novo assembly of a metagenome typically yield a collection of contigs that cannot readily be grouped by species. Methods for generating chromatin-level contact probability maps, e.g., as generat ...[more]

PMID: 24855317

Similar Datasets

Project description:Over the last decade, structure-function relationships have begun to encompass networks of brain areas rather than individual structures. For example, corticostriatal circuits have been associated with sensorimotor, limbic, and cognitive information processing, and damage to these circuits has been shown to produce unique behavioral outcomes in Autism, Parkinson's Disease, Schizophrenia and healthy ageing. However, it remains an open question how abnormal or absent connectivity can be detected at the individual level. Here, we provide a method for clustering gross morphological structures into subregions with unique functional connectivity fingerprints, and generate network probability maps usable as a baseline to compare individual cases against. We used connectivity metrics derived from resting-state fMRI (N = 100), in conjunction with hierarchical clustering methods, to parcellate the striatum into functionally distinct clusters. We identified three highly reproducible striatal subregions, across both hemispheres and in an independent replication dataset (N = 100) (dice-similarity values 0.40-1.00). Each striatal seed region resulted in a highly reproducible distinct connectivity fingerprint: the putamen showed predominant connectivity with cortical and cerebellar sensorimotor and language processing areas; the ventromedial striatum cluster had a distinct limbic connectivity pattern; the caudate showed predominant connectivity with the thalamus, frontal and occipital areas, and the cerebellum. Our corticostriatal probability maps agree with existing connectivity data in humans and non-human primates, and showed a high degree of replication. We believe that these maps offer an efficient tool to further advance hypothesis driven research and provide important guidance when investigating deviant connectivity in neurological patient populations suffering from e.g., stroke or cerebral palsy. Hum Brain Mapp 38:1478-1491, 2017. © 2016 Wiley Periodicals, Inc.

Project description:PurposeDeep-learning-based segmentation models implicitly learn to predict the presence of a structure based on its overall prominence in the training dataset. This phenomenon is observed and accounted for in deep-learning applications such as natural language processing but is often neglected in segmentation literature. The purpose of this work is to demonstrate the significance of class imbalance in deep-learning-based segmentation and recommend tuning of the neural network optimization objective.MethodsAn architecture and training procedure were chosen to represent common models in anatomical segmentation. A family of 5-block 2D U-Nets were independently trained to segment 10 structures from the Cancer Imaging Archive's Head-Neck-Radiomics-HN1 dataset. We identify the optimal threshold for our models according to their Dice score on the validation datasets and consider perturbations about the optimum. A measure of structure prominence in segmentation datasets is defined, and its impact on the optimal threshold is analyzed. Finally, we consider the use of a 2D Dice objective in addition to binary cross entropy.ResultsWe observe significant decreases in perceived model performance with conventional 0.5-thresholding. Perturbations of as little as ±0.05 about the optimum threshold induce a median reduction in Dice score of 11.8% for our models. There is statistical evidence to suggest a weak correlation between training dataset prominence and optimal threshold (Pearson r=0.92 and p≈10-4 ). We find that network optimization with respect to the 2D Dice score itself significantly reduces variability due to thresholding but does not unequivocally create the best segmentation models when assessed with distance-based segmentation metrics.ConclusionOur results suggest that those practicing deep-learning-based contouring should consider their postprocessing procedures as a potential avenue for improved performance. For intensity-based postprocessing, we recommend a mixed objective function consisting of the traditional binary cross entropy along with the 2D Dice score.

Dataset Information

Species-level deconvolution of metagenome assemblies with Hi-C-based contact probability maps.

Publications

Species-level deconvolution of metagenome assemblies with Hi-C-based contact probability maps.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets