Association Of Genes/Pathways To Phenotypes

Question

[Pi]

Greetings

I was hoping to obtain some recommendations of available resources to find genes/pathways associated with a phenotype (not necessarily human).

I have some general phenotype information but I do not know which disease(s) it is associated with. I also have sequence data for the affected individuals. As a general starting point I was hoping to look for genes and pathways that might be associated with this phenotype. I was then going to link this knowledge to an analysis of any variations in the individuals. I am very receptive to alternative approaches though :)

I have seen OMIM and OMIA. I have also seen PhenomicDB and PhenoHM: human-mouse comparative phenome-genome server.

I am aware there is a myriad of GWAS resources linking markers to phenotypes but I was looking specifically for resources linking genes and pathways to phenotype. I have chosen (perhaps unwisely) not to focus too much on the GWAS studies because the marker identified by the study may not be the causal/functional SNP. I do have QTL data to narrow down my genomic regions of interest so hopefully that will compensate.

Please can you advise if I have missed any obvious useful resources or if there are any other strategies I could employ to get a starting point.

Thank you for your time

Answer 1

[Larry_Parnell]

A very relevant question - but back at you: What is a phenotype?

Classically, this was seed color or shape or some other easily visible and measured quantity. Height, body weight, waist circumference are ones we use in obesity research. Eye color and blood type are other good examples.

Could a phenotype be a disease? Certainly. In this case, mine OMIM for genes associated with diseases. KEGG also has some disease pathways. There is an excellent paper by Zhang, Becker, et al (2010) on 1462 human genes affiliated with disease based on a comparison (and substantiation) of human and mouse phenotypes associated with those genes. They assign genes to 480 different diseases.

Could mRNA levels be a phenotype? You bet, and so we have eQTL - expression quantitative trait loci. There's a BioStar thread on eQTL databases. Or enhancer activity: See Wasserman's excellent paper on prostate cancer and a TP53 (p53) enhancer element altered by the disease-associating SNP. A phenotype could be interaction with a microRNA - see Brest et al.

So, the list of phenotypes goes on. GWAS are a source - one that offers a lot. But you can and should look for phenotype associations in many other places. OMIM is one. The mouse and rat genome data repositories are others.

Answer 2

[Chris Evelo]

I am afraid a lot of it is likely based on GWAS, but [?]gen2phen.org[?] should be an interesting resource in this respect. [?]The human variome project[?] also wants to collect this kind of data, but currently they mainly have links to pilot projects. [?]Leiden Open Variome Databases[?] were designed for this purpose. They have a list of current instances, but I am not sure which ones are open access, if any.

[Chris Evelo]

You are absolutely right! Getting the phenotype and environmental factors (diet to start with) is one of the hardest problems. It is good to see that those larger initiatives now are at least starting up, but it will take time to collect the genotype/phenotype/environment relationships. I am always surprised on the other hand how much of that kind of information is presented on posters at conferences and never gets published. We need people with mobile devices going to those conferences and collecting that kind.

[Pi]

Thanks for your reply. I have looked at LSBDs but there aren't any relevant for my scenario. I'm also very familiar with the gen2phen project. I was really hoping for more resources along the line of PhenomicDB where I can search for a particular phenotype and get genes found to be involved in this phenotype. Is there a reason why there aren't many resources of this nature? Normally in bioinformatics you can find a plethora of resources with similar data/functionality and the dilemma is which one to use. I was surprised at the paucity of resources mapping known phenotypes to genes and pathways

Answer 3

[David Quigley]

You could look at the Mouse Genome Informatics (JAX) web site, specifically the Phenotype and Allele query form. For example, you can query a phenotype like "hyperlipidemia" and get back a list of QTL loci and targeted mutants (all in mouse models, naturally) that are associated with the phenotype. They also host a Mammalian Phenotype ontology browser you may find useful; for example, here's a result set of alleles associated with "heart left ventricle hypertrophy".

Answer 4

[Khader Shameer]

I am not aware of any computational resource/method that map phenotype to pathways without using genotype information, so in this case you can't ignore GWAS data.

I have tried to explain some of my thoughts about the limitations in inference of phenotype/genotype association with pathways in previous discussions (see here and here). Also see the figure 4 in review article here, and an update figure here. Current survey of association study indicates that most of the variants are in the non-coding, gene-desert regions. Only few examples were validated like the 9p21 and another example on association of genetic variants with expression of proximal genes here.

As you are working in non-human data, I would recommend you to make use of orthology searches using available human association studies here. I am not aware of any other resource that map phenotype to pathways. I will recommend to use the available human GWAS data, filter the data using some conditions like (replicated in >2 studies, appropriate OR and P-values) and if your phenotype or related phenotype exists (click Open/Close Phenotype Tree) in Association Results Browser, get the associated genes and perform a bi-directional best-hit blast search against your genome of interest. Another option is to look into eQTL databases for human orthologs. As you indicated in your question if you have QTL information, that will be definitely helpful.

From a computational perspective you can do an orthology based QTL searches. Again select your phenotype from association browser, get associated genes, perform search in databases like SCANDB and perform ortholog search against your genome to compare the results.

[Pi]

Hello. I have seen those biostar questions and it raises some excellent points (in fact I've previously quoted LPs example in that question about the dangers of assuming that because a SNP is in a certain gene that it is affecting the p/w the gene participates in). My problem is that I'm coming from the problem from an opposite angle to these questions. I'm not trying to map a SNP to a disease pathway. I have a phenotype from an unknown disease and I'm trying to find which of my millions of SNPs could be responsible

[Pi]

Regarding the non-coding gene desert point, if a single SNP in a gene desert caused a phenotype through some sort of regulatory mechanism, would you still see mendelian inheritance of the phenotype if you had pedigree data just as you would for a mutation in a single gene?

[Pi]

Regarding the non-coding gene desert point, if a single SNP in a gene desert caused a phenotype through some sort of regulatory mechanism, would you still see mendelian inheritance of the phenotype if you had pedigree data just as you would for a mutation in a single gene? I don't think I've seen an example of this.

[Pi]

I'm afraid I've never worked with GWAS data before so this is probably a basic question. My understanding is that the TAS isn't necessarily the causal SNP but the TAS could well be in linkage disequilibrium with the causal SNP. I had discounted GWAS because, as you say, most SNPs are in gene deserts and I only have exome data. I was assuming I'd be unlikely to find any exonic SNPs in LD with trait associated SNPs in gene deserts and intergenic regions. Was that a fair assumption?

[Pi]

That said, i had't seen the association results browser before and that certainly makes life a lot easier!