A new study in which Sergi Valverde, from the Complex Systems Lab at DCEXS-UPF has collaborated, provides an open source MATLAB package to study the structure of bipartite ecological networks inspired by real problems in microbiology and with broader applications.
Bipartite networks are a special type of ecological network where individuals of a certain species interact with individuals of different species. They are ubiquitous in community ecology, such as the relation between phages (viruses that infect bacteria) and their bacterial hosts.
You can read more about this tool, called BiMat, at the group’s blog.
Very interesting talk by Edward Marcotte today at the PRBB!
He is an expert in proteomics, but touches all aspects of systems biology, and today he asked the following question: how does genotype determine phenotype? Can we predict the outcome of all the genomic variation we are uncovering with the many genomic projects we are doing nowadays?
Well, his lab is certainly trying to do so, and using three different strategies which I will summarise very briefly:
1. Using functional gene networks, which are based on data such as mRNA expression, protein-protein interactions (PPI), etc. These networks presumably are formed by genes that are involved in the same biological processes. From here one only needs to follow the “guilt-by-assotiation” principle and assume that, if a gene in that network is involved in a particular phenotype (a disease, for example), the genes around it might also be so. They have tested this in yeast, C.elegans, Arabidopsis, rice and mouse, at least. They have managed to validate predictions for up to 200 genes. And they have come up with a valuable principle: that phenotypes reflect biological modules, rather than single proteins. That is, it is not a specific proteins that is essential, but a specific complex.
2. A systematic mapping of stable protein complexes in humans, which they have done in collaboration with labs in Toronto and which includes more than 2000 Mass spec experiments. From here they have inferred more than 600 stable complexes in humans (more than 500 of them with more than 3 components), of which 1/3 are unknown. Now the idea is to use this PPI network as a framework for linking genes to diseases. And they are doing so with one children developmental disease, the Cornelia de Lange syndrome, for which 3 known genes explain only the 50% of cases. They have selected some of the proteins which are around those three in the network and are currently sequencing them in patients.
3. Using model organisms to infer human disease genes. This is by far the one that I was most surprised about. It turns out that looking for what he called phenologs (orthologous phenotypes between organisms, for example, which yeast phenotype is equivalent to breast cancer in humans) one can find surprising disease models. For example, yeast sensitivity to lovastatin is a model for angiogenesis defects in humans! This is found looking for the yeast orthologous of human genes involved in angiogenesis, and checking which phenotype those yeast genes are involved in. Then one can look at the rest of the yeast genes involved in that phenotype, and check if their human orthologous might be involved in angiogenesis.
And then, in principle, one could even use screening in yeast to find angiogenesis inhibitors. And the surprising thing is that it works! The Marcotte lab is actually about to start a phase I clinical trial on a drug they found this way and which they hope might be useful for glioblastoma. This is just one example, but according to him, this ‘phenologs’ strategy seems to work for more than 50% of the human genetic diseases… One big lesson that stems from this knowledge is that protein modules are conserved through evolution even if the phenotype is not – a concept he called ‘evolutionary repurposing’. Very interesting indeed.
Report by Maruxa Martinez, Scientific Editor at the PRBB