Annotation consistency : ChIP experiments
Caution must be used when annotating results of an in vivo ChIP experiment for a protein-DNA interaction due to the uncertainty of whether that interaction is direct or not.
However, Collas (2010) review describes suitable controls required for ChIP and points out that ChIP has become the technique of choice for mapping DNA–protein interactions in the cell, identifying novel binding sites for transcription factors or other chromatin-associated proteins, mapping the localization of post-translationally modified histones and mapping the localization of histone variants.
From Rachael (GOA);
In an initial email exchange with Sylvain Poux from SIB, he said he would use caution when annotating ChIP experiments because "A ChIP experiment does not indicate whether interaction is direct or not since chromatin complexes are all cross-linked with DNA, they only show colocalization. "
I contacted two curators at the IntAct protein interaction database (http://www.ebi.ac.uk/intact/main.xhtml) to get their opinions on this and these are their responses;
Sandra Orchard; "We do annotate ChIP data, though with varying degrees of enthusiasm - Dave Thorneycroft and I both accepted it, Jyoti is not a fan. However, I would agree with the statement about not showing a direct interaction. At the moment it goes in as a Physical Association, which indicates a reasonably tight complex - we could also consider a global update to the more vague term "Association" which is reserved for affinity complexes and indicates that it comes down as a messy complex with potentially all sorts of interactors in there. That is probably a more realistic view of the situation. I think colocalisation is not as appropriate since the PCR analysis is usually pretty specific as to which gene this affinity complex is binding to. We do not capture the data when its only been shown to bind to a non-specific transactivation site and not a specific gene. I would hesitate to infer a protein:DNA binding an in vivo ChIP - Sylvain is quite correct in pointing out that there could be a lot of other material between your protein bait and DNA prey.
You do occasionally find in vitro expts where only purified protein and a strand of DNA are involved - those you can be a lot more positive about but unfortunately these are comparatively rare."
Jyoti Khadake; "Apart from some of the transcription factors and histones most other proteins that are normally found in a chip assay are ancilary proteins that can at best be called structural or modulating proteins. Often these have little if any contact with the DNA itself. Unless there is a gel retardation assay or footprinting with purified proteins it is very difficult to assess the interaction of the protein in a complex with DNA. You could have a Go term 'DNA interacting complex' and then the term would be valid here. Often in these experiments only the proteins of interest are looked at this further compounds the issue of not detecting the correct DNA interactor. The preparation in vivo is as dirty as any cytoskeletal pull down. I guess colocalize will need further clarification. Colocalization to nucleus would be silly really since most chromatin is in nucleus. However most of the ChIP isolated proteins can be shown to be in the same pixel on immuno-microscopy and electron tomography and often on electron-microscopy as well."
Sandra Orchard; We only annotate ChIP if there is definitely a cross-linker involved so you can have some degree of confidence that the protein is at least indirectly associating with the nucleic acid, and we only use our vaguest interaction type 'association', to imply we know there will be a lot of other proteins also present. Native ChIP, i.e. where there is no cross-linker used, we do not capture as it's often unclear whether they are purifying a true complex, or artefactually exposing new DNA regions to proteins during the purification process. The different ChIP methodologies (as detailed in "The Current State of Chromatin Immunoprecipitation" by Philippe Collas, PMID:20077036) tend to refer more to how they identify the components i.e. qPCR, NGS etc. which is irrelevant to the actual attacment of the protein to DNA (or not). We also only capture this data if there is a clear identification of gene or nucleic acid sequence - we ignore it if its just generic, unspecified DNA so you could also insist with an 'interacts with' being supplied.
Update 5 Nov 2014 Discussed via [SourceForge] In summary: An acceptable annotation based on ChIP is protein x "chromatin" (or a child term) and then include Sequence Ontology (SO) term in the annotation extension field.
Note that the definition of chromatin is: The ordered and organized complex of DNA, protein, and sometimes RNA, that forms the chromosome therefore it is implied that any protein identified by ChIP is part of the chromatin rather than binding the chromatin.