Dataset Information

Shallow Sparsely-Connected Autoencoders for Gene Set Projection.

ABSTRACT: When analyzing biological data, it can be helpful to consider gene sets, or predefined groups of biologically related genes. Methods exist for identifying gene sets that are differential between conditions, but large public datasets from consortium projects and single-cell RNA-Sequencing have opened the door for gene set analysis using more sophisticated machine learning techniques, such as autoencoders and variational autoencoders. We present shallow sparsely-connected autoencoders (SSCAs) and variational autoencoders (SSCVAs) as tools for projecting gene-level data onto gene sets. We tested these approaches on single-cell RNA-Sequencing data from blood cells and on RNA-Sequencing data from breast cancer patients. Both SSCA and SSCVA can recover known biological features from these datasets and the SSCVA method often outperforms SSCA (and six existing gene set scoring algorithms) on classification and prediction tasks.

SUBMITTER: Gold MP

PROVIDER: S-EPMC6417803 | biostudies-literature | 2019

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Shallow Sparsely-Connected Autoencoders for Gene Set Projection.

Gold Maxwell P MP LeNail Alexander A Fraenkel Ernest E

Pacific Symposium on Biocomputing. Pacific Symposium on Biocomputing 20190101

When analyzing biological data, it can be helpful to consider gene sets, or predefined groups of biologically related genes. Methods exist for identifying gene sets that are differential between conditions, but large public datasets from consortium projects and single-cell RNA-Sequencing have opened the door for gene set analysis using more sophisticated machine learning techniques, such as autoencoders and variational autoencoders. We present shallow sparsely-connected autoencoders (SSCAs) and ...[more]

PMID: 30963076

Similar Datasets

Project description:The use of bioinformatic tools for read-based taxonomic and functional analyses of metagenomic data sets, including their assembly and management, is rather fragmentary due to the absence of an accepted gold standard. Moreover, most currently available software tools need input of millions of reads and rely on approximations in data analysis in order to reduce computing times. These issues result in suboptimal results in terms of accuracy, sensitivity, and specificity when used either for the reconstruction of taxonomic or functional profiles through read analysis or analysis of genomes reconstructed by metagenomic assembly. Moreover, the recent introduction of novel DNA sequencing technologies that generate long reads, such as Nanopore and PacBio, represent a valuable data resource that still suffers from a lack of dedicated tools to perform integrated hybrid analysis alongside short read data. In order to overcome these limitations, here we describe a comprehensive bioinformatic platform, METAnnotatorX2, aimed at providing an optimized user-friendly resource which maximizes output quality, while also allowing user-specific adaptation of the pipeline and straightforward integrated analysis of both short and long read data. To further improve performance quality and accuracy of taxonomic assignment of reads and contigs, custom preprocessed and taxonomically revised genomic databases for viruses, prokaryotes, and various eukaryotes were developed. The performance of METAnnotatorX2 was tested by analysis of artificial data sets encompassing viral, archaeal, bacterial, and eukaryotic (fungal) sequence reads that simulate different biological matrices. Moreover, real biological samples were employed to validate in silico results. IMPORTANCE We developed a novel tool, i.e., METAnnotatorX2, that includes a number of new advanced features for analysis of deep and shallow metagenomic data sets and is accompanied by (regularly updated) customized databases for archaea, bacteria, fungi, protists, and viruses. Both software and databases were developed so as to maximize sensitivity and specificity while including support for shallow metagenomic data sets. Through extensive tests performed on Illumina and Nanopore artificial data sets, we demonstrated the high performance of the software to not only extract taxonomic and functional information from sequence reads but also to assemble and process genomes from metagenomic data. The robustness of these functionalities was validated using "real-life" data sets obtained from Illumina and Nanopore sequencing of biological samples. Furthermore, the performance of METAnnotatorX2 was compared to other available software tools for analysis of shotgun metagenomics data.

Project description:Rural communities often rely on groundwater for potable water supply. In this study, untreated groundwater samples from 28 shallow groundwater wells in Finland (<10 m deep and mostly supplying untreated groundwater to <200 users in rural areas) were assessed for physicochemical water quality, stable water isotopes, microbial water quality indicators, host-specific microbial source tracking (MST) markers, and bacterial community composition, activity, and diversity (using amplicon sequencing of the 16S rRNA gene and 16S rRNA). Indications of surface water intrusion were identified in five wells, and these indications were found to be negatively correlated, overall, with bacterial alpha diversity (based on amplicon sequencing of the 16S rRNA gene). High levels of turbidity, heterotrophs, and iron compromised water quality in two wells, with values up to 2.98 nephelometric turbidity units (NTU), 16,000 CFU/ml, and 2,300 μg/liter, respectively. Coliform bacteria and general fecal indicator Bacteroidales bacteria (GenBac3) were detected in 14 and 10 wells, respectively (albeit mostly at low levels), and correlations were identified between microbial, physicochemical, and environmental parameters, which may indicate impacts from nearby land use (e.g., agriculture, surface water, road salt used for deicing). Our results show that although water quality was generally adequate in most of the studied wells, the continued safe use of these wells should not be taken for granted. IMPORTANCE Standard physicochemical water quality analyses and microbial indicator analyses leave much of the (largely uncultured) complexity of groundwater microbial communities unexplored. This study combined these standard methods with additional analyses of stable water isotopes, bacterial community data, and environmental data about the surrounding areas to investigate the associations between physicochemical and microbial properties of 28 shallow groundwater wells in Finland. We detected impaired groundwater quality in some wells, identified potential land use impacts, and revealed indications of surface water intrusion which were negatively correlated with bacterial alpha diversity. The potential influence of surface water intrusion on groundwater wells and their bacterial communities is of particular interest and warrants further investigation because surface water intrusion has previously been linked to groundwater contamination, which is the primary cause of waterborne outbreaks in the Nordic region and one of the major causes in the United States and Canada.

Dataset Information

Shallow Sparsely-Connected Autoencoders for Gene Set Projection.

Publications

Shallow Sparsely-Connected Autoencoders for Gene Set Projection.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure