Young Bioinformatician Award

Annamaria Necsulea is awarded for her work on "The evolution of vertebrate tissue transcriptomes". Anamaria Necsulea carried out the research she is awarded for at the Functional Evolutionary Genomics group led by Henrik Kaessmann at the University of Lausanne. She currently is an Ambizione fellow (SNSF) at EPFL.

Complex organisms depend on the presence of a wide variety of cell types and tissues to function. Despite their striking physiological and morphological differences, the genetic material is essentially the same. What changes is the way the genetic instructions are interpreted: gene activity not only differs between cells and tissues but also during the lifetime of an organism. In the same way, many morphological and physiological differences between species can be explained by evolutionary changes in gene activity, rather than by differences in gene structure or gene sequence.

To understand these issues, Anamaria recently focused on a particular class of genes: the long non-coding RNAs (lncRNAs). lncRNAs do not encode proteins but are believed to play an important part in the cell, such as modifying the level of activity of other genes. Unlike protein-coding genes, lncRNAs have not been studied in detail. Consequently, their precise functions, modes of interaction with other genes and contribution to phenotypic traits are still largely unknown.
With this in mind, Anamaria set out to investigate the lncRNA repertoires of 11 tetrapod species and performed the first large-scale evolutionary analysis of this class of genes. She identified ~11,000 primate-specific lncRNA families and ~2,400 highly conserved lncRNAs, including ~400 genes whose origin probably dates back 300 million years. She discovered that lncRNAs – in particular ancient ones – are actively regulated and seem to function predominantly in embryonic development. Importantly, she compared expression patterns of homologous lncRNA and protein-coding families across species with an end to reconstruct an evolutionarily conserved co-expression network. This led to an initial understanding of the roles of lncRNAs in cellular expression networks on a genome-wide scale.

The network, which surprisingly contains many lncRNA hubs, suggests potential functions for lncRNAs in fundamental processes like spermatogenesis or synaptic transmission, but also in more specific mechanisms such as the control of placenta development through miRNA production.

SIB Best Graduate Paper Award

This year we have two co-winners of the SIB Best Graduate Paper Award. They share the cash prize of CHF 5'000.

Slavica Dimitrieva is a PhD student in the Computational Cancer Genomics Group led by Philipp Bucher at EPFL, Lausanne.

In the past years, Slavica has been studying the vertebrate ultraconserved non-coding elements (UCNEs) which are one of the biggest mysteries in current biology. UCNEs are almost 100% identical between species that are as distant as human and shark. To this day, however, there is no known molecular mechanism that requires such a high degree of conservation. Experiments suggest that most of the UCNEs act as transcriptional regulators of master developmental genes.

The work of Slavica followed up on the observation that UCNEs cluster around key developmental genes. With Phillip Bucher, they sought to understand whether this strong co-localization reflects functional cooperativity between these elements or, alternatively, if each element acts independently on the target gene, in which case the clustering would merely reflect the importance of the target gene.

Their work revealed the striking fact that UCNEs that are part of a regulatory region are either jointly retained or jointly lost after the whole genome duplication in teleost fishes. Such a result can only be explained by a high degree of connectivity of the UCNEs in an interaction network. Another very exciting aspect is that this result could explain the ultraconservation; if one UCNE interacts with many others then each interaction will impose constraints on the base sequence.

The results from this work can be explored via the public resource UCNEbase and read in ‘Genomic context analysis reveals dense interaction network between vertebrate ultra-conserved non-coding elements’ published in Bioinformatics, September 2012.

Charles Vejnar did his PhD under the supervision of Evgeny Zdobnov at the University of Geneva, and is now a Postdoc at Yale University (USA) in Giraldez Lab.

In animals, the production of proteins starts with the transcription of protein-coding genes to messenger RNAs (mRNAs), which are then translated into proteins. This process is complex and strongly regulated to produce the required amount of proteins linked to the tissue identity. While transcription factors regulate the expression of genes at the transcription level, microRNAs (miRNAs) induce a post-transcriptional repression of protein-coding genes. The repression starts by reducing translation followed by mRNA degradation. Embedded in the RNA-Induced Silencing Complex (RISC), miRNAs act as a recognition element for driving the RISC to repress targeted mRNAs. The rules guiding this recognition are however not fully known: partial base-pairing between the start of the miRNA (5' end, the so-called “seed”), and the targeted mRNA in its 3’-untranslated region (UTR) is known to trigger repression of mRNA expression. However, with such a rule, millions of potential miRNA targets can be identified to the ~1000 human miRNA genes.

Prioritization of targets for any miRNA functional analysis is therefore of critical importance to study miRNA function, and requires the ranking of potential miRNA targets bearing a seed-match. We proposed to predict a biologically meaningful ranking criterion: the effect on mRNA or protein levels of miRNA-mediated repression. We developed an open source software library – iRmap – which, for the first time, covers a wide range of approaches to predict miRNA repression. MiRmap includes 13 features covering thermodynamic, evolutionary, probabilistic, and sequence-based approaches. For example, miRmap novel features measure the significance of the target site’s evolutionary conservation, and the statistical significance of seed occurrence(s) in 3’-UTR sequences using an exact probability distribution.

The predictive power of miRmap features was evaluated in an unbiased way using high throughput experimental data from immunopurification, transcriptomics, proteomics and polysome fractionation experiments. This wide range of experiments allowed us to evaluate miRmap performance on the different aspects of miRNA-mediated repression such as target recognition, mRNA stability and effects on translation. Overall, target site accessibility to the miRNA RISC is the most predictive feature. MiRmap model combining all features almost doubles the predictive power of the renowned TargetScan tool by increasing the proportion of variance explained from 7.5% to 13% of miRNA over-expression effects measured at the transcriptome level. Prediction features were also tested with experimental data, obtained on tissue samples instead of cell line cultures in collaboration with Swiss research groups. In particular, we investigated the role of miR-122 in the regulation of the hepatocyte transcriptome, of miR-155 in dendritic cell maturation and function, and the effects on transcriptome of knocking-out all miRNAs in mouse testis.

Available as an open source Python library and as a web application (, miRmap establishes a solid foundation for the future development of approaches to miRNA target prediction, facilitating meaningful comparisons between existing and new features, and providing the community with direct access to state-of-the-art analytical tools.