Twin Studies and Challenges with Mixed Cell Populations (2/4/2015)

Wed, Feb 4, 2015 at 11:13 AM

Customer: (1) How is your company typically involved in the process of DNA-methylation and RNA-seq analysis? Do you typically help with experimental design or do most clients just send you the data they have already collected? Since the design of these study impacts the quality of the data, it is important to keep the analysis in mind while designing the experiments.

(2) What is the typical turnaround time for analysis of, for example, an RNA-seq data set? A typical experimental design for us would involve comparing RNA-seq results from a set patients with a disease to those without the disease, with the goal identifying gene expression differences that could serve as novel biomarkers or provide mechanistic insight into the disease.

(3) Do your bioinformatics specialists and statisticians have any experience working with twin studies? We have a cohort of twins for which we are interested in methylation and RNA-sequencing analysis, and hope to maximize the amount of additional statistical information we can acquire based on a twin study design (e.g. heritability analyses).

(4) We deal mostly with mixed populations of cells. Clearly sorting pure populations of cells is ideal, but not always possible. Do your bioinformatics specialists and statisticians have experience with and techniques designed for analyzing data from mixed cell populations?

Wed, Feb 4, 2015 at 4:06 PM

AccuraScience LB: (1) Most researchers approach us after they have completed the experiments and need the data analyzed. Indeed, in some cases, while analyzing/interpreting the data, we wish that the researchers had consulted us before-hand and the studies had been designed differently. We would be happy to work with you when the study is still being designed.

(2) Turn-around time for RNA-seq data analysis (including sequencing data quality control, mapping of reads to the reference genome allowing exon-exon junctions to be identified, expression quantification, differential expression analysis, and pathway analysis to identify over-represented biological pathways among differential expressed genes) is about 4-5 weeks. We always try to accommodate the researcher's schedule, however. Thus if you are trying to meet a deadline for a grant application, for example, please let us know and we will adjust our priorities and push your project faster.

(3) Data from twin studies are fairly straight-forward to analyze from statistics point of view. In most cases a linear mixed effect model would do. We have a lot of experience with this kind of work.

(4) Fairly "advanced" mathematical modeling methods have been developed to distinguish, or "deconvolve" compositions of cancer samples which are known to be a mixture of multiple subpopulations (or subclones). But those methods take advantage of the different mutation signatures of the subclones which are revealed in the sequencing data. For non-cancer samples, those methods will not apply. For non-tumor samples that are a mixture of multiple cell types (or tissue types), it is theoretically not possible to deconvolve the compositions by examining the sequencing data alone. If the sample impurity issue is bothering you, I would suggest that you look into some recently developed "single-cell RNA-seq" techniques. They take advantage of microfluidics or other technologies to pick individual cells, and profile their transcritomes one at a time. These methods are not without limitations, however - do a Google search if you are interested.