Biology subjects

Winther, O.

Publications and source records attributed to Winther, O..

21 records · Page 2Linked to original sources

bioRxiv2018

Binning microbial genomes using deep learning

Identification and reconstruction of microbial species from metagenomics wide genome sequencing data is an important and challenging task. Current existing approaches rely on gene or contig co-abundance information across multiple samples and k-mer composition information in the sequences. Here we use recent advances in deep learning to develop an algorithm that uses variational autoencoders to encode co-abundance and compositional information prior to clustering. We show that the deep network is able to integrate these two heterogeneous datasets without any prior knowledge and that our method outperforms existing state-of-the-art by reconstructing 1.8 - 8 times more highly precise and complete genome bins from three different benchmark datasets. Additionally, we apply our method to a gene catalogue of almost 10 million genes and 1,270 samples from the human gut microbiome. Here we are able to cluster 1.3 - 1.8 million extra genes and reconstruct 117 - 246 more highly precise and complete bins of which 70 bins were completely new compared to previous methods. Our method Variational Autoencoders for Metagenomic Binning (VAMB) is freely available at: https://github.com/jakobnissen/vamb

bioinformatics↗

bioRxiv2018

scVAE: Variational auto-encoders for single-cell gene expression data

MotivationModels for analysing and making relevant biological inferences from massive amounts of complex single-cell transcriptomic data typically require several individual data-processing steps, each with their own set of hyperparameter choices. With deep generative models one can work directly with count data, make likelihood-based model comparison, learn a latent representation of the cells and capture more of the variability in different cell populations.\n\nResultsWe propose a novel method based on variational auto-encoders (VAEs) for analysis of single-cell RNA sequencing (scRNA-seq) data. It avoids data preprocessing by using raw count data as input and can robustly estimate the expected gene expression levels and a latent representation for each cell. We tested several count likelihood functions and a variant of the VAE that has a priori clustering in the latent space. We show for several scRNA-seq data sets that our method outperforms recently proposed scRNA-seq methods in clustering cells and that the resulting clusters reflect cell types.\n\nAvailability and implementationOur method, called scVAE, is implemented in Python using the TensorFlow machine-learning library, and it is freely available at https://github.com/scvae/scvae.

bioinformatics↗

bioRxiv2018

NetSurfP-2.0: improved prediction of protein structural features by integrated deep learning

The ability to predict local structural features of a protein from the primary sequence is of paramount importance for unravelling its function in absence of experimental structural information. Two main factors affect the utility of potential prediction tools: their accuracy must enable extraction of reliable structural information on the proteins of interest, and their runtime must be low to keep pace with sequencing data being generated at a constantly increasing speed.\n\nHere, we present an updated and extended version of the NetSurfP tool (http://www.cbs.dtu.dk/services/NetSurfP-2.0/), that can predict the most important local structural features with unprecedented accuracy and runtime. NetSurfP-2.0 is sequence-based and uses an architecture composed of convolutional and long short-term memory neural networks trained on solved protein structures. Using a single integrated model, NetSurfP-2.0 predicts solvent accessibility, secondary structure, structural disorder, and backbone dihedral angles for each residue of the input sequences.\n\nWe assessed the accuracy of NetSurfP-2.0 on several independent test datasets and found it to consistently produce state-of-the-art predictions for each of its output features. We observe a correlation of 80% between predictions and experimental data for solvent accessibility, and a precision of 85% on secondary structure 3-class predictions. In addition to improved accuracy, the processing time has been optimized to allow predicting more than 1,000 proteins in less than 2 hours, and complete proteomes in less than 1 day.

bioinformatics↗