Proteins are the workhorses of the cell and, over billions of years, they have evolved an amazing plethora of extremely diverse and versatile structures with equally diverse functions.Therefore, protein evolution echoes the evolution of all forms of life.While it was previously assumed that all proteins result from old ones and that evolutionary emergence of new proteins and transitions between existing ones are rare or even impossible, recent advances in comparative genomics have repeatedly called some 10%-30% of all genes without any detectable similarity to existing proteins.Even after careful scrutiny, some of those "orphan" genes contain protein coding reading frames with detectable transcription and translation but lack a homolog at the DNA level in outgroups.Accordingly, some proteins seem to have emerged from previously non-coding 'dark genomic matter'.Computational studies indicate that these 'de novo' proteins are disordered, fast evolving, weakly expressed and rapidly assuming novel and physiologically important functions.They are therefore interesting to understand…

The field of Artificial Intelligence has made significant progress over the past 5 years by making use of Deep Neural Networks. The ability of Deep Neural Networks to carry out complex mappings in high dimensional spaces makes them extremely interesting for biological applications, e.g. for relating genotypes to (molecular) phenotypes. This talk gives an easy to understand introduction into Deep Learning concepts that are interesting for Life Sciences. State of the art results are presented by using Deep Neural networks for predicting DNA binding motifs and RNA secondary structures.

Alternative splicing is a major gene expression regulatory process that increases the diversity of the transcriptome and the proteome encoded by a limited number of genes. For example, 95% of human genes generates different splicing variants encoding different protein isoforms. While tremendous progress has been made in elucidating the molecular mechanisms of alternative splicing regulation (relying on RNA biding proteins or splicing factors), two important related issues have been poorly addressed. The first issue concerns the biological information encoded by coregulated alternative exons (e.g., exons regulated by the same splicing factor). To date, it has been assumed that splicing factor-coregulated exons are diverse and unrelated in terms of coding information features. A second major issue concerns the relationship between coding and exonic splicing regulatory sequences. So far, it has been assumed that coding sequences can accommodate splicing regulatory sequences because of the redundancy of the genetic code (a same amino acid is…

Vaccination has saved more lives than any other medical procedure. But, some pathogens have evolved which have defied successful vaccination using the empirical paradigms pioneered by Pasteur and Jenner. I will describe how bringing together theoretical/computational approaches from statistical physics, with immunology is beginning to confront this challenge by developing some of the principles necessary for rational design of vaccines that may eliminate these scourges. One characteristic of many pathogens for which successful vaccines do not exist is that they present themselves in various guises. HIV is an extreme example because of its high mutability, and it continues to wreak havoc, especially in developing countries. I will describe how we developed models to translate data on HIV protein sequences to knowledge of the HIV fitness landscape, and tested the resulting predictions against in vitro and clinical data (with collaborators). Based on these studies, a therapeutic T cell-based vaccine was designed, which is…

The co-transcriptional assembly of mRNAs into messenger ribonucleoparticles (mRNPs) is pivotal for both genome expression and integrity: indeed, the combinatorial association of RNA-binding proteins (RBPs) to mRNAs does not only define their cellular fate, but also counteracts the accumulation of genotoxic DNA:RNA hybrids, or R-loops. To further understand how these processes are coordinated, especially for functionally related mRNAs, we systematically analyzed genome-wide maps of RBP association and R-loop formation in the budding yeast S. cerevisiae. On the one hand, we have investigated whether specific mRNP biogenesis pathways could take place for mRNAs encoding the subunits of multiprotein complexes. Focusing on the Nuclear Pore, we highlighted a trans-acting mechanism that collectively target the mRNAs encoding the components of this complex, and prevent the accumulation of unassembled subunits. On the other hand, we have searched for cis-acting genomic features regulating R-loop formation. Surprisingly, we found that the presence of introns mitigates the accumulation of…

Considered as the dark matter of genomes, the non-protein coding DNA portion of eukaryotic genomes has, in the last decade, been shown to be much bigger than what was thought. Furthermore, with the advent of deep sequencing it became clear that genomes are pervasively transcribed and that the noncoding fraction generates huge amounts of transcripts that cannot be accounted for as simple “junk” or transcriptional noise. This challenges the central dogma of molecular biology where, for over 50 years, proteins have been considered to be the sole drivers of biological processes. Recent evidence strongly supports that noncoding (nc)RNAs are indeed functional and responsible for a multitude of regulatory processes such as gene regulation, nucleus organization and post-transcriptional regulation. In Phaeodactylum tricornutum over ∼1,500 long intergenic (li)ncRNAs and ∼3,300 natural antisense lncRNAs (NATs) have been identified. These are polyadenylated bona fide discrete transcripts that are expressed under specific physiological states. A ‘guilt by…

The co-evolutive landmarks left by the evolutionary process in protein multiple sequence alignments can be used, among other things, to detect pairs of proteins physically interacting or functionally related. Over time, we have developed different approaches that exploit this idea with the aim of extracting protein interaction networks from the deluge of sequence data that characterizes modern Biology. Current “high-throughput sequencing” technologies have the potential to revolutionize these methods as they generate protein sequences in a fast and cheap way. Nevertheless, more than blindly using all available sequences, it looks like the future will be characterized by a careful selection of those to use, and even the “a-la-carte” re-sequencing of the more suitable species/strains for a particular family/system of interest.