Animal Health & Welfare

Analysis of "next generation sequencing" data

The emergence and spread of anthelmintic resistance in parasitic nematodes is a serious threat to the sustainability of the sheep industry in Scotland. Unfortunately, mechanisms of resistance are still poorly understood. "Next generation sequencing" technologies offer an opportunity to detect the changes in genetic activity that take place in response to anthelmintic treatment. Analysis of this new class of data presents new statistical challenges.

Studies at MRI into a triple drug-resistant isolate of the nematode Teladorsagia circumcincta using Roche 454 sequencing have generated 100,000 short subsequences of cDNA from both ivermectin-exposed and unexposed groups. These subsets of sequences were assembled into contiguous sequences, BLAST-identified and clustered by putative function. The number of observed subsequences per cluster in each of the anthelmintic exposed and unexposed groups quantifies the level of gene activity associated with that cluster. However, cost issues mean that only a single observation is available from each combination of cluster and treatment.

In the absence of replication, our approach has been to categorise the clusters with respect to the p-values calculated from two statistical models, each of which we know models the data inappropriately, but which together span the range of results which might conceivably arise from a more appropriate, but inestimable, model. Fitting generalised linear models to the numbers of observed sequences in each cluster, we fit liberal and conservative models, and categorise each cluster according to the statistical significance of treatment differences under the two alternative models. Significance under the conservative model indicates the strongest evidence for a treatment effect, allowing scientists to prioritise their follow-up activities.

barley plots picture   barley plots picture

The nematode Teladorsagia circumcincta is an important parasite of sheep.

 

Histogram of cluster classifications as a function of p-values from the conservative model.

     
     

Further details from: Mintu Nath

Article date 2009

Consultancy Advice & Collaboration

Plant Science

Animal Health & Welfare

Ecology & Environmental Science

Human Health & Nutrition