Probabilistic Functional Networks

February 8, 2008

I recently read the Lee et. al. paper “A single gene network accurately predicts phenotypic effects of gene perturbation in Caenorhabditis elegans” in Nature Genetics. This is an extension of their previous work on building probabilistic functional networks to a multicellular organism (the worm). (Most of their prior work was in yeast.)

A probabilistic functional gene network attempts to detect functional links between proteins by integrating large systematically generated datasets [1]. In this view of a gene, pleiotropy is implicit. It is, in a sense, an automatic method of generating ontologies using the guilt by association paradigm. Each dataset is weighted based on its ability to recapitulate benchmarks (such as GO categories; curated knowns; etc).

In previous work the Marcotte lab generated a probabilistic functional network for Saccharomyces cerevisiae [2], improved that network by reducing bias and noise in the original construction [3], and showed that the connectivity of a probabilistic functional network can predict loss-of-function phenotypes [4]. A nice example of how, given the right perspective, one can use the power of statistics to identify weak but significant signal among noisy diverse datasets.

The extension to Caenorhabditis elegans sought to address a lingering question: could this sort of integrated network methodology accurately predict gene function in multicellular animals? After all, cell types, tissues, and developmental stages are not explicitly reflected in the network. They construct a network (Wormnet v1) covering approximately 82% of the worm’s proteome. They then demonstrate the utility of this network to make predictions which are cell specific. They first identify genes that function in the retinoblastoma tumor suppressor pathway (affecting differentiation of three vulval precursor cells). They then examine a predicted connection between the dystrophin associated protein complex (DAPC) and the EGF signaling pathway, conclusively showing that DAPC positively regulates EGF-Ras-MAPK signaling (in the single worm excretory cell).

These probabilistic functional networks are static summaries which integrate information from mutants, different cell types, tissues, varied conditions, developmental time, and space. Consequently, they are expected to have certain limitations. Nevertheless, the probabilistic functional network paradigm has proven to be a powerful method for abstracting functional/informational linkages between genes in both a single celled model system (S. cerevisiae) and now a more complex multicellular model system (C. elegans). Wormnet can be queried through an online interface at

So there are two obvious directions for this work. First, as more data becomes available it can be readily integrated into the framework — improving the network and (perhaps) converging on a more complete description of the system. Second, throughout the worm paper they make claims about the feasiblity of this approach to tackling more complicated mammalian systems such as human and/or mouse. Signals of things to come from this group.

  1. Fraser, A.G., Marcotte, E.M. (2004). A probabilistic view of gene function. Nature Genetics, 36(6), 559-564. DOI: 10.1038/ng1370
  2. Lee, I. (2004). A Probabilistic Functional Network of Yeast Genes. Science, 306(5701), 1555-1558. DOI: 10.1126/science.1099511
  3. Lee, I., Li, Z., Marcotte, E.M., Califano, A. (2007). An Improved, Bias-Reduced Probabilistic Functional Gene Network of Baker’s Yeast, Saccharomyces cerevisiae. PLoS ONE, 2(10), e988. DOI: 10.1371/journal.pone.0000988
  4. McGary, K.L., Lee, I., Marcotte, E.M. (2007). Broad network-based predictability of Saccharomyces cerevisiae gene loss-of-function phenotypes. Genome Biology, 8(12), R258. DOI: 10.1186/gb-2007-8-12-r258

Lee, I., Lehner, B., Crombie, C., Wong, W., Fraser, A.G., Marcotte, E.M. (2008). A single gene network accurately predicts phenotypic effects of gene perturbation in Caenorhabditis elegans. Nature Genetics, 40(2), 181-188. DOI: 10.1038/ng.2007.70


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: