Association analysis rely on historical recombination events, so, unlike linkage analysis, you don't need pedigree information or controlled crosses to identify genes contributing to a trait. Also, because it assumes individuals are unrelated or have little population structure, any significant marker-trait associations would be tightly linked, narrowing the interval(s) within the linkage group where the causative gene or genes can be mapped

It would be difficult compare both: Genetic association studies and linkage analysis. Both of them are important in terms of their application. For mapping disease loci in Mendelian disorders in families, we perform genetic linkage analysis. Where as in multifactorial or complex disorders, both environment and genetic play role in disease manifestation and to analyze the role of genetics in them, we perform genetic association studies by analyzing disease group and ethnically matched normal control samples. I hope it explains the importance of both the techniques and clarification of your question as well.

Greatly appreciating all these valuable answers from all you seniors. Cheers.

While linkage studies are best suited to monogenic disorders they can still be applied to multifactorial disease locus mapping and they have been widely used for multiple complex disorders.

The main differences between the two methods is that Linkage studies are based on linkage analysis in families with a particular phenotype segregating and they usually have poor genetic resolution in the centimorgan range while association studies are based on population studies in affected subjects that detect linkage disequilibrium with fine genetic resolution in the kilobase range.

In practical terms the big advantage of association studies over linkage studies is that you do not need family studies you just need subjects with a well defined phenotype. The other advantage is that linkage works on a broad genetic scale so it is possible to get a positive linage result for a large region of DNA which may contain several or even many genes of small effect that may be difficult to tie down when the region is examined more closely. If you find an association with a SNP you know that you are either in the gene responsible or there is another neighbouring gene in strong disequilibrium with your SNP.

There is also a way to do combined linkage and association for complex disorders when you have cases and parents (TDT or FBAT). In these cases, you look at the probability of transmitting a risk allele - 50% under null, >50% to cases (50% to unaffecteds if you have them as well). It may guard more agains population stratification, and does not have the big problem of exactly matching cases and controls on a lot of measures since the same people, the parents, are used for "caseness" and "control" - the transmitted alleles are the case chromosomes, the nontransmitted are the control chromosomes. For childhood disorders it would be advantageous. Anyway else to chime in why this is not so commonly used?

Very very valuable answers keep coming, greatly appreciating.

I just want to add some more to what have been mentioned already:

Dealing with complex disease is difficult, and linkage analysis has strong limitations. Imagine something like Diabetes or Skeletal deformities, they are multi-factorial. To do a proper linkage analysis not only you need a family of multiple generations or several families with the same disorder, you also should be able to define a phenotype clearly. What is going to be your cut off range of blood sugar or the angel of skeletal curvature to distinguish your case from your control specifically? how we can include the very mild versions of these disease (which means they have the genes but may be not all the required variants). By excluding the so called healthy grandparents who never have been checked carefully for Scoliosis before they die and considering them healthy in your pedigree, are not you going to be miss leaded?

But when it comes to population studies since you are not looking for a specific inheritance pattern, free of any hypothesis you will come up with the SNPs or variants that could be the cause or at least be responsible for the severity of disease.

Cheers greatly appreciated

One of the problems with association analyses is collecting a group of patients with a clearly defined phenotype. This is somewhat easier to control for within a family or a consanguineous population. When you have a group of related individuals, you can perform linkage analyses. In the end, the results of the linkage analysis on the complex trait will indicate which regions of the genome contribute to the trait in that particular population but will not describe all of the genetic variation responsible to the trait in the general population. The results you get from the linkage analyses will inform your ideas regarding the metabolic pathways, developmental lineages and cell types involved.

Very much appreciated.

To add to the general discussion here. There is a further advantage to association studies in comparison to linkage when dealing with polygenic disorders at a population level. In linkage work, unless you have a VERY large family you need to combine data from multiple families. For polygenic disorders it is highly unlikely that every family will have the same collection of underlying genetic factors and so linkage results will conflict and your LOD scores flatten. And it only takes a small proportion of you families to not map to the same disease loci for a highly informative LOD map to 'dissolve'. On the other hand, in association studies the individual association between the test phenotype and particular genomic regions will be weakened if multiple genetic predispositions exist, but will still be detected. So GWAS is likely to always be a more powerful method for detecting genes associated with multifactorial disease. On the other hand, doing GWAS in human populations you need always bear in mind that many 'populations' consist of several ethnic groups who have only recently (less than 10 generations) admixed. In these sorts of populations care must be taken with the use of case:control comparisons since it is possible for two ethnic groups to have different genetic SNP frequencies and disease prevelance. In some cases you then 'pick up' a group of associations at above chance levels that are really markers of shared population history prior to admixture between the affected individuals in the mixed population rather than shared descent from an affected ancestor. So you can get false positives with association analysis in substratified populations if you are not careful, which can be avoided in linkage analysis.

I am new to GWAS, although I have used homozygosity mapping and linkage analysis is Saudi population where 80% consanguinity exists. I have families with multiple affected siblings with unknown inheritance pattern suffering from T1Diabetes. Could you please suggest, if I should  procEed with Linkage analysis or GWAS. 

Dear Atia Shaheen, Microarray along with whole genome/exome sequencing will be quite useful in analyzing your families. You have not mentioned about the outcome of the homozygosity mapping. If you have already mapped the the families to certain candidate loci, you can go ahead with targeted exome sequencing, otherwise Microarray in combination whole genome sequencing will yield the desired results. I hope my suggestion would serve the purpose. Please feel free to contact for any further discussions.

I am writing a proposal on families with multiple affected siblings with Type 1 diabetes and I have not done any experiment yet. As these families are usually consanguineous, I am confused if I should go with GWAS or linkage analysis as the first step and later pull out the rare variant using Whole exome sequencing or whole genome sequencing.