(Dec. 8th, 2009) A new "-ome" is around. Besides the well-established "genome", "transcriptome" or "metabolome", the “negatome” now comes as a sort of antagonist to the “interactome” -- it lists pairs of proteins or domains that, with high probability, do NOT physically interact with each other. Kathleen Gransalke reports.
Publishing negative data was confirmed by the Journal of Negative Results in Biomedicine in the editorial of their first issue back in 2002: “[…] we would like to draw the reader's attention to [famous science philosopher] Karl Popper's realization that science advances through a process of ‘conjectures and refutations’. Popper gave a rather compelling and simple example: for thousands of years Europeans believed that swans are white based on observations of millions of white swans, until exploration of Australasia introduced Europeans to black swans. Popper's point: only one black swan was needed to repudiate the theory that all swans are white. However, many confirming instances there are for a theory, it only takes one counter observation to falsify it.”
The negatome database described by Smialowski et al. in the latest issue of Nucleic Acid Research could now provide equally useful information as protein interaction studies often lead to false positive or false negative results; the reason for this being the high noise in many experimental setups. Furthermore, the database could turn out to be a real time-saver because if two proteins are proven unlikely to interact, you don’t have to do the experiment at all and can simply move on (if, of course, you trust the database)! As an example, the negatome can tell you that mouse synaptotagmin-like protein 5 doesn't interact with Rab-18, despite the fact that it binds very well to other Rab proteins.
Two complementary datasets contain all non-interacting protein pairs or NIPs found so far by Smialowski et al. The first set comprises manually curated data and the second one is based on structural information provided by the Protein Data Bank (PDB). In the manually curated dataset, negative results are collected from mammalian NIPs extracted from 246 scientific articles. Accumulating this data proved to be rather difficult because negative results don’t usually get published; thus, most of the data comes from control experiments or whenever several proteins from one family were tested for interaction with the target protein. Altogether, 1,291 pairs made it into the list.
In the second set, experimental structures of biological units provided by the PDB were analysed for inter-chain distances between Cβ atoms. When the distance was too large, a pair of protein chains was declared non-interacting. This set also includes non-mammalian proteins and contains 809 pairs. Both sets were filtered against the IntAct database of known interaction pairs; in addition to it, pairs of known interacting PFAM domains were subtracted using the 3DID and iPFAM databases.
What you have in the end is a rather accurate set of protein pairs that most likely do not interact with each other. So, before you get your hands dirty trying to find possible protein interactions, you might want to check the negatome database first from now on for the likelihood of a protein-protein interaction.
The database, funded by the Biosapiens Network of Excellence, is freely accessible via the website of the Munich Information Center for Protein Sequences at the Helmholtz Zentrum München.