May 7-8, 2018, hosted by:

Bielefeld University, Germany Room X-B2-103

Map of University Campus here. (We are in building X.)

Map of X Building here. (We are in section B, floor 2, room 103.)

by Alessandro Mammana

For some doctoral candidates, postdoctoral education and scientific research may appear as the natural continuation of their PhD in bioinformatics. For others, leaving the academic world behind and working in a company may look like a better opportunity. This talk is about my career path, which brought me from a PhD in bioinformatics to a job as a bioinformatics scientist working at Illumina and currently to a job as software engineer working at Google. I will discuss what I like about research and what I like about industry, the lessons that I learned and the opportunities that the PhD opened up for me.

by Alexander Sczyrba

Methods for assembly, taxonomic profiling and binning are key to interpreting metagenome data, but a lack of consensus about benchmarking complicates performance assessment. The Critical Assessment of Metagenome Interpretation (CAMI) challenge has engaged the global developer community to benchmark their programs on complex data sets, generated from ~700 newly sequenced microorganisms and ~600 novel viruses and plasmids and representing common experimental setups. In my talk I will present results from the first CAMI challenge for assembly, binning and profiling tools. These results highlight current limitations of metagenomic tools, but also provide a roadmap for software selection to answer specific research questions.

by Michael Sammeth

Ten years ago, RNA-seq turned up as a revolutionary tool for monitoring the transcriptome compositions of arbitrary cell types and states, opening up many novel possibilities for studies on functional genomics. This also gave rise to several multi-national projects to conduct RNA-Seq at large scale in different human cells, in order to advance with our understanding of how the genomic information influences our phenotypes, even after death. Furthermore, the combined RNA-sequencing of host cells together with pathogens (“Dual-Seq”) has recently demonstrated the capacity to monitor the cross-talk between two organisms during an infection. Functional studies in non-model organisms, however, are usually complicated primarily by their incomplete genomic and transcriptomic information. In this talk, we will discuss different possibilities to support RNA-seq studies by bioinformatics paradigms, and we discover some interesting biology along our way.

by Barbara Hammer

Feature selection methods aim for an automated identification of the most relevant characteristics for a given regularity based on given examples alone such as the automated identification of the potentially most relevant biomarkers in a biomedical diagnosis task. While a plethora of methods exists which can efficiently identify clear signals, the situation turns out more complicated in the realm of feature correlations and feature redundancies: on the one hand, it is not possible to exhaustively test subsets of features due to the involved combinatorial explosion; on the other hand, it is unclear how to interpret found feature relevances if features are potentially redundant hence the identified features are not unique.

In the talk, I will present a metric learning technology which can identify relevant sets of features via an adaptive feature relevance learning scheme which is integrated into the machine learning classifier. On the other hand, I will discuss effects which can occur in the case of redundant features, i.e. relevant features sets are not unique, and the results of standard feature selection technologies can be highly misleading. I present the so-called all relevant feature selection problem, and present novel machine learning technology how to solve this exemplary biomedical applications.