Microarrays vs. next-gen sequencing II

Saturday, November 22th, 2008 by hinrich

Next-gen sequencing

Theoretically, next-gen sequencing should be much more accurate in quantifying the number of transcripts present in a given sample for a certain sequence than microarrays. This is also the focus of attention in many articles and advertisements highlighting the benefits of next-gen sequencing for quantitative assessments of gene expression alterations (also referred to as RNA-seq). The main concept is: Transfrom mRNA into cDNA and sequence everything. Furthermore, as no hybridization takes place, issues such as cross-hybridization or biases introduced in the design of probes are avoided.

However, after the initial hype and the sale of many instruments of the major technology providers, scientists start to realize the following: Yes, next-gen sequencing can be very accurate in sequencing DNA and one can quantify very accurately how many copies of a given sequence were present in a sample. But no, all the laboratory steps prior to the actual sequencing done by the next-gen machines are not free of introducing biases. In other words, some sequences are much more easily transformed into cDNA than others (destroying the linearity necessary to conclude that the number of sequences read are representative of the number of copies that were present in the sample originally). Furthermore, some platforms require enzymatic processing steps (e.g. generation of cDNA libraries using random primers) that also introduce biases into the cDNA sequences. And these "wrong" sequences are accurately read by the sequencers.

While the potential of the next-gen sequencing machines is very encouraging, the laboratory procedures are far from perfect. Similar to the years of research that went into determining the strengths and weaknesses of microarray technology, similar efforts will need to go into next-gen sequencing so that scientists are able to discriminate between sequences that were truely present in the sample vs. those that were "generated" during the preparation of the sequencing sample. This is especially an issue with large genomes (in contrast to e.g. yeast). Therefore, next-gen sequencing is likely to focus the coming years on creating content rather than being used for quantitative analysis of trancript abundances.

How about a study similar to the MAQC study conducted by the FDA to scientifically assess how quantitative the different next-gen sequencing platforms really are for gene expression studies?

Posted in Molecular Profiling

Using the FARMS approach for copy number data

Friday, November 07th, 2008 by hinrich


While our I/NI-approach seems to be very useful for the removal of uninformative genes when studying gene expression data, the detection of changes in the number of DNA copies along a given chromosome is still not optimal. Here the same problem exists: are potential alterations in the copy number state due to a true change or is one only looking at false positive changes caused by too much noise in the data?

We are currently working on an approach (copynumber.FARMS or cn.FARMS) to utilize similar concepts of the gene expression world (the FARMS approach and the I/NI-calls) for the detection of DNA copy number changes (CNVs). In contrast to many other approaches, it seems like we can get away without needing to correct for effects caused by the GC content of the probes on the Affymetrix SNP arrays.

Okko has just presented his work using 500k and SNP 6.0 data at the 11th International Meeting of MGED. The R code will be made available on the web pages of the bioinformatics institute of the Johannes Kepler university, Linz: Software

Posted in Molecular Profiling

Almost finished

Monday, November 03th, 2008 by hinrich

Art work to be used for the book

We have received the feedback of the reviewer of Chapman & Hall for the book Willem and I are currently working on ("Gene Expression Studies Using Affymetrix Microarrays"). Bottom line of the reviewer was: "I've reviewed the book and I like the structure and philosophy, although I do have some critiques". That was encouraging! I personally like the last sentence of the reviewer's general comments best: "There's much to like here and much that would be useful in general not just in the context of microarrays".

Therefore we are doing our best to come up with a manuscript that will be worth the money and the time a potential reader would invest. Deadline is 1st of December for submitting the final text. Folks from Chapman & Hall will then do the proofreading while we will continue to polish the figures and the layout. Taking some more time to incorporate the suggested changes, we should be able to finish the book sometime in January/February next year.

Posted in Molecular Profiling