Milí kolegové a kolegyně,
chtěli bychom vás srdečně pozvat na další přednášku semináře, která se koná ve středu 3.5.2017, od 17:20 v posluchárně S4.
Hostem na semináři bude Karel Sedlář z Ústavu biomedicínského inženýrství na VUT v Brně a téma přednášky bude
Visualizations in metagenomics
Těšíme se srdečně na vaši účast!
Za všechny organizátory semináře s pozdravem, Petr Daněček
Stránky semináře http://bioinformatika.mff.cuni.cz/seminar/
--------------------------------------
Karel Sedlář Visualizations in metagenomics
One of main steps in a study of microbial communities is resolving their composition, diversity and function. In the past, these issues were mostly addressed by the use of amplicon sequencing of a target gene because of reasonable price and easier computational postprocessing of the bioinformatics data. With the advancement of sequencing techniques, also whole metagenome shotgun sequencing became popular, allowing much more detailed analysis of the metagenomics data, including reconstruction of novel microbial genomes and to gain knowledge about genetic potential and metabolic capacities of whole environments. Clear and easy-to-understand visualization plays an important role within both of these sequencing approaches. Amplicon sequencing of a barcode sequence, e.g. bacterial 16S rRNA gene, provides an easy way to perform qualitative, as well as quantitative analysis of a microbial community. A wide range of techniques can be used to visualize datasets represented by relative abundances of OTUs (Operational Taxonomic Units) in different samples. Dimensionality reduction techniques, e.g. PCoA (Principal Coordinate Analysis) using UniFrac metrics, provides informative visualization showing relations between samples without a need for identification of OTUs against a reference database. Bipartite graphs and other network visualizations, on the other hand, highlight the differences and similarities among samples by showing unique and shared taxa. At last, not at least, clustered heatmaps of correlation coefficient can show relations among microbiome composition and other parameters, e.g. presence of antibiotic resistance genes. Visualization in whole metagenome shotgun sequencing approach often accompanies clustering of sequences, so called binning. Since lot of techniques perform clustering in unsupervised manner, the use of interactive inspection and visualization can be suitable for validity verification of the binning output or even for manual fine-tune of the automatic clustering. The visualization methods use information regarding sequence composition or coverage to produce coordinates in two- or three-dimensional space to describe the given sequence. In some cases, the sequences can be visualized simply by the use of the parameters, e.g. OFDEG (Oligonucleotide Frequency Derived Error Gradient), GC content and others, as coordinates without any further transformation. In the majority of cases, however, the sequences are described by more than three parameters, disabling them from being projected into a humanly comprehensible space. Denouement is then brought by use of dimensionality reduction techniques, such as PCA (Principal Component Analysis), SOM (Self-Organizing Maps), or t-SNE (t-distributed Stochastic Neighbor Embedding).