Annotating from phenotypes
Mutant phenotypes can provide important insights into gene function. In order to use mutant phenotypes as supporting evidence for a GO annotation, however, curators should keep in mind the following general principles:
- GO annotations based on a mutant phenotype should represent the normal role of a gene in biology.
- Mutant phenotypes need to be interpreted in the overall context of what is known about a gene so that appropriate Biological Process terms can be selected; some phenotypes represent manifestations of genetic perturbations far downstream from the actual function of a gene.
- GO does not aim to capture individual phenotypes. Most groups that contribute annotations to GO also curate phenotypes as a separate data type, allowing GO curators to be selective about phenotypes used for annotation knowing that more complete and detailed phenotype curation is often captured elsewhere.
- The guidelines for annotating from phenotypes refer to several relations from the Relations Ontology (RO).
Gene/Gene Product to GO Term Relations
- involved in
- Formal definition: C involved_in p if and only if c enables some process p', and p' is part of p
- For any given Biological Process, think of the Molecular Functions that make up that process and using that information, decide if the gene/gene product being annotated enables one of those functions. If so, then the gene/gene product is 'involved in' that process.
Ask yourself: what is the normal Biological Process for this gene?
- Remember that annotations based on mutant phenotypes are inferences about the normal Biological Process (BP) for a gene.
- To decide if a gene is directly 'involved in' a BP, determine if its Molecular Function (MF) is one of the MFs that makes up that BP.
- To help with this, it may be useful to create an activity-based GO-CAM model in Noctua or consult a pathway or process model from a recent paper or review.
- If the MF of a gene is not known, curators must be very careful about making a BP annotation based solely on a phenotype.
- For genes with unknown MFs, in the absence of additional analysis (e.g. genetic and physical interactions), phenotypes can generally only be annotated to a BP using the ‘acts upstream of or within’ relation (or one of its positive or negative effect child relations). In these cases, it is also perfectly acceptable to not make a GO annotation until more is known about the gene.
Being ‘required for’ a process does not mean a protein is ‘involved in’ a process
- In GO annotation, making a statement that a gene is 'involved in' a process is an explicit statement that the MF of that gene is an integral part of the BP.
- Although authors may state that a gene is 'required for' a process, GO curators must interpret those statements within the defined semantics of a GO annotation and decide if that means the gene's activity is truly a part of the BP.
- If the MF is not part of the BP, consider using a less specific relation between the gene and the BP, or not annotating for BP until more information is known.
Pleiotropic effects need to be considered carefully
- Pleiotropy, in which a single variation can result in multiple defects, needs to be carefully considered in the context of GO BP annotation.
- An understanding of the molecular mechanism underlying pleiotropic defects can help determine if the gene is integrally involved in each process, or affects a process more generally required.
Beware of read-outs
- Biologists often use experimental "read-outs" to assess a gene's role in a larger biological process.
- The affected "read-out", however, is not always indicative of the process in which the gene is most proximally involved.
- For example, DNA fragmentation is often used as a "read-out" for apoptosis, but mutations that result in increased DNA fragmentation do not necessarily indicate that the corresponding gene is directly involved in DNA catabolism.
- Again, consider the MF of the mutated gene and how it fits into the BP used as a "read-out" as well as the overall BP being studied.
- Some papers may describe how genomic perturbations affect relatively 'high-level' processes. Examples of 'high-level' processes include:
- Aging or determination of adult lifespan
- Cell death
- Cell division
- Cell migration - examples? Cancer disease states? Metastasis? Look at co-occurrence of migration and proliferation (and others in this list).
- Cell proliferation
- Development (embryonic, larval)
- Protein localization
- Reproduction and brood size
- If the mechanism underlying the phenotypic effects on these processes is not known, we strongly encourage curators to not make annotations to the corresponding BP term.
- Creating a GO-CAM model or consulting a recent paper or review can help determine the basis for the phenotype and whether the BP annotation is appropriate, and if so, with what relation.
- Mutually exclusive terms: cell proliferation should not be used for proteins involved generally in growth or division. Examples?
Mutant phenotypes and regulation terms
- Mutant phenotypes may more often be used as supporting, rather than definitive, evidence that a gene is involved in regulation of a process.
- The GO BP ontology contains terms for processes as well as regulation of processes.
- From mutant phenotypes alone, it can be difficult to assess whether a gene is involved in a process or regulates a process.
- As with annotating to BP terms from phenotypes more generally, consider what is known about the MFs involved in the process and use that information to decide if annotating to the process or regulation of the process is best.
- To annotate to regulation of the process, there should be evidence that the MF of the gene product being annotated regulates an MF of a gene that is involved in the process.
- Mutants annotated to ‘regulation’ with no molecular function annotation (sometimes an annotation to a protein complex of with a known function) should be examined closely and reviewed.
A gene is not "involved in" a process, but there is still a biologically meaningful annotation I want to capture
- Sometimes, despite the fact that the relation between a gene and a BP is not clear, a curator may wish to capture the association.
- This situation may occur when there is otherwise little else known about a gene.
- In these cases, curators may make a BP annotation based on a phenotype using the most general annotation relation, acts upstream of or within, or one of its positive or negative effect child terms.
- If a gene is known to act upstream of a BP, curators may also consider using the acts upstream of relation or one of its positive or negative effect child terms.
- In these cases, though, it is important to think about how relevant and informative the resulting 'acts upstream of' annotation would be for GO users.
- Consider making a GO-CAM model in this case to help delineate how closely 'upstream' the gene's activity really is.
Reviewing and removing older phenotype-based annotations
- When there is new knowledge, older IMP annotation should be reviewed and removed if they no longer accurately reflect the relationship between a gene's MF and its role in a BP.
- In these cases, GO annotations reflect the summation of knowledge about that gene, while phenotype annotations (not GO annotations) will continue to reflect what has been experimentally observed for mutations in that gene.
- Q: Is it possible to have a gene product with a novel or unknown MF (or just a protein binding MF) that is otherwise well characterized genetically (and maybe somewhat biochemically?) such that it could be placed within a given pathway or process? Does anyone have an example of this, where the annotation would be informative?
- Example: nuclear pore: BRR6 is involved in nuclear envelope organization, when mutated, causes nucleocytoplasmic transport defects, but is NOT involved in nuclear transport
- An essential nuclear envelope integral membrane protein, Brr6p, required for nuclear transport (2001). (PMID:11483521). The authors identified an essential gene, BRR6, whose product, Brr6p, behaves like a nuclear envelope integral membrane protein. Notably, the brr6-1 mutant specifically affects transport of mRNA and a protein reporter containing a nuclear export signal. In addition, Brr6p depletion alters nucleoporin distribution and nuclear envelope morphology, suggesting that the protein is required for the spatial organization of nuclear pores.
- Integral membrane proteins Brr6 and Apq12 link assembly of the nuclear pore complex to lipid homeostasis in the endoplasmic reticulum (2010). (PMID:20016074)
The authors found that Brr6 has an essential function in regulating lipid homeostasis in the NE-ER, thereby impacting NPC formation and nucleocytoplasmic transport.
- GO:0055088 lipid homeostasis
- GO:0006998 nuclear envelope organization
- NOT nucleocytoplasmic transport.
It is common for authors to state that a gene or gene product is required for a given process. However, the true meaning of this statement can vary and it is thus the responsibility of the curator to determine whether the gene or gene product's activity is, indeed, an integral part of a process. To help make this decision, it may be useful to think about the gene's role in a concentric circle of processes, starting with its MF. For a given MF, what is the most proximal process in which the gene is involved? For a transcription factor, the most proximal process would be regulation of transcription. From there, what is the next most proximal process? In a development context, for example, this might be cell fate specification or differentiation. Moving outward, what phenotypes might be indicative of defects in cell fate specification? If, for example, the cell was a particular class of neurons, then one of the defects may manifest as a change in a behavior. Is the transcription factor thus 'involved in' the behavior? No, but by annotating the TF to regulation of transcription (perhaps with relevant target genes) and specification of the particular cell types, it is possible capture the most relevant aspects of that gene's function. The defects in behavior could then be captured with phenotype annotations. Possible example: C. elegans ttx-1
apoE example? https://www.uniprot.org/uniprot/P02649
For example: splicing factors are often required for cell cycle transition, but they are not part of the cell cycle transition A good clue is viability of mutants: inviable mutants often have pleiotropic phenotypes or they have a strong terminal phenotype that can easily be misinterpreted (cell cycle transition blocks/checkpoints, chromosome mis-segregation, etc)