It is more than a week since ASHG 2013 ended. Among the many interesting sessions and talks over four days of fun, here is a quick summary of a few of the talks that stood the “Remember me after a week” test. There are already a few nice summaries of ASHG 2013 by @massgenomics and @erlichya lab. After the first timer’s view on ASHG, this is a computation/transcriptome biased view of ASHG 2013.
One of the talks which I did not even attend, but left a great impression (thanks to tweets), was Jeanne B Lawrence’s fascinating work on using a single gene to silence a whole down syndrome chromosome 21. Should I even name the gene? Well. it is XIST, the X-inactive specific transcript that silences one of the X chromosomes. Down Syndrome is caused by trisomy for chromosome 21. In a beautiful work published in Nature earlier in the summer this year, Lawrence’s team asked whether one can use XIST and silence one of the three chromosome 21 and showed that they can do it in iPS cells derived from a Down’s syndrome patient.
ASHG Future: Mo Tissue, Mo Transcriptomics, and Mo Single Cell Transcriptomics
Other than Mo GWAS loci and Mo Data, one thing that stood out at ASHG 2013 is in my subjective view was Mo tissue, Mo Transcriptomics, Mo Single Cell Transcriptomics. Probably it was GTEx effect :) GTEx had about a dozen talks plus posters. But, this is for sure a sign of things to expect in the future. Multi-tissue, Multi-datatype and single cell sequencing are here to stay.
One of the interesting talks was by Borel on transcriptomes of individual cells. As one expects gene expression showed individual-cell specific signatures. The team also looked at allele-specific expression at individual cell.
In brief, we diploids have two copies of every gene (autosomal), where one copy came from mother and the other copy came from father. Allele-specific expression is preferential expression of one copy over the other, i.e. mother’s copy (allele) over father’s copy and vice versa.
Typically in bulk, a lot of genes show preferential expression of one alleles over the other, but both the alleles are expressed (See results from two largest transcriptome sequencing efforts published recently). Very interestingly, the team found that the allele-specific expression at individual cell showed typically only one allele is expressed.
It is kind of understandable for low expressed genes, as there be may just one or few copies of mRNA in a cell (and those could completely biased toward one of the alleles). But, for highly expressed genes, the idea of only one of maternal/paternal allele expression is the norm rather than exception is very intriguing and points to interesting biology.
At the same time one has got to be careful over interpreting the result as it could be due to some technical artifact. Unfortunately, caution is warranted as single cell transcriptomics is at a very similar stage as RNA-seq technology during the time of finding of excessive (and all types of) RNA-editing. Let us keep the fingers crossed till then.
It was nice to see very nice discussion with similar thoughts on twitter almost immediately after the talk started by Tuuli Lappalainen. (Click on Retweets to see full discussion.)
One cell, one expressed allele. What happens w heterozyg variants with severe recessive cellular effects? Do 50% of cells die?
— Tuuli Lappalainen (@tuuliel) October 24, 2013
On the computational side a couple of interesting talks kind of addressed a theme/question with the future in mind. The theme is “What can we do with millions of genome/exome sequences?”
Monkol Lek from Daniel MacArthur’s group addressed the computational challenges with using over 60,000 exomes to call variants and the benefit of using all the data in variant calling. Needless to say, dealing with over 60K exome data is mind boggling. The team had to reduce the original bam files of size to 750TB to just 50TB. Massively Impressive. It will be really interesting to know more details on computational side of the project in detail.
Iqbal from McVean’s Group addressed an interesting and challenging question “how can we move away from our over-simplisitic view of genome as a linear sequence A,T,G,and C. Even though sequencing efforts have characterized genetic variations, our linear reference genome model does not lend itself to utilize the known variations in any downstream sequence analysis easily. McVean team had extended the simple reference genome to “reference genome graph” that allows easy representation of genetic variations and sequence similarity. Using the reference genome graph approach, the team has used the 1000 Genomes data to build a human population level reference graph. The idea of variations/similarity aware genome representation is exciting and points to a lot interesting applications.
These are just samplers, there were many more interesting sessions/talks. ASHG 2013 abstract book is there to summarize them all :)