Two interesting papers related to our research October 30, 2008

Pat Brown (the inventor of microarrays) just published a paper in PLoS Biology in which he described the identification of mRNA targets for 40 RNA-binding proteins. Not only it’s a very interesting study that emphasizes the extent of post-transcriptional regulation, but they also extensively use FIRE for motif discovery ( along with their homebrewed approach)

Hogan DJ, Riordan DP, Gerber AP, Herschlag D, Brown PO (2008) Diverse RNA-Binding Proteins Interact with Functionally Related Sets of RNAs, Suggesting an Extensive Regulatory System. PLoS Biol 6(10): e255

The other paper in Nature by Ruschlow and colleagues describes a protein that binds to a putative reglatory element (CAGGTAG) that we identified in our collaboration with Eric Wieschaus as highly over-represented upstream of the earliest zygotic genes in the Drosophila embryo (De Renzis, Elemento et al, PLoS Biology, 2007).They used one-hybrid assay to identify the zinc finger protein called Zelda.

Liang HL, Nien CY, Liu HY, Metzstein MM, Kirov N, Rushlow C. The zinc-finger protein Zelda is a key activator of the early zygotic genome in Drosophila. Nature.

Postdoctoral positions in computational genomics at Weill Cornell Medical College August 11, 2008

Here’s the ad. A computational biologist will already join my group in the fall (jointly hired with another group at Weill Cornell that produces massive amounts of genomic data).

Olivier Elemento’s computational biology group at the Weill Medical College of
Cornell University in Manhattan, New York, has several openings for postdoctoral scientists.
The group’s research emphasis is on human regulatory genomics and
proteomics. We develop innovative computational tools and approaches to
generate testable hypotheses and discover fundamental principles in
transcriptional, post-transcriptional and post-translational regulation of gene
expression.
Our research is done in close collaboration with experimentalists at Weill and
elsewhere and our ultimate goal is to develop translational research towards
novel understanding and therapy for diseases such as cancer and
neurodegenerative disorders.

For more details on the group’s research, please consult :

http://physiology.med.cornell.edu/faculty/elemento/lab/

Strong analytical and programming expertise is required. Preference will be given
to individuals with a proven track record in computational genomics. Individuals
with strong computational skills and experimental background are also
encouraged to apply.
To apply, send a resume in addition to names and contact addresses of two
persons who can provide recommendation letters to ole2001@med.cornell.edu

And if you join my group, this is the kind of projects you could be working on.

G Protein-Coupled Receptors July 29, 2008

Being about to join a medical school, I have a growing interest in G protein coupled receptors (GPCRs). The reason is that 30% of the 800 known GPCRs are drug targets, and some of them are the targets of multiple drugs (half of the drugs target a GPCR, according to an interesting article in the current issue of The Scientist). GPCRs appear to control many processes, eg appetite, smell, heartbeat, sleep, etc. Some are involved in development (e.g. GPR50-null mutant mice never begin puberty), while at least another one can mediate retroviral entry (eg CCR5, targeted by anti-HIV Maraviroc). GPCRs usually bind a ligand, but some of them simply block other GPCRs by dimerizing with them (eg GPR50 blocks MT1 this way).

Could computational/systems biology help understand how GPCRs work ? Maybe. I am curious to see what the GPCR co-expression network (eg across human tissues) is like and how that correlates with drug targeting, whether that can predict heterodimerization or drug side effects. Since there is a lot of data about transcriptional response to drug treatment (eg, with the Connectivity Map from the Broad), it seems to me that it would also be interesting to ask how the transcriptional program upon treatment with drugs targeting different GPCRs actually differ (and whether there is some overlap). Also I am curious to see where GPCRs are expressed in the brain. If anybody has any other ideas, or is interested in working on some of these problems, please shoot me an email.

Moving to Weill Medical College of Cornell University July 25, 2008

On September 1st, I will be starting my own group at the Weill Medical College of Cornell University in Manhattan. My lab will be located at the Institute for Computational Biomedicine (ICB).

I have many exciting projects in mind, some of them will build on FIRE, FIRE-pro, PAGE and FastCompare, others are completely new. The overall goal will be to understand and predict cellular activity and regulation based on sequence (DNA, RNA and protein). The ultimate goal will be to develop translational research towards understanding diseases such as cancer, neurodegenerative disorders (and I am particularly interested in investigating brain gene expression in order to study the latter) and malaria. Importantly, I will seek close collaborations with experimentalists, and in fact I already have started collaborating with some experimental groups there at Weill.

I also have funds to hire several postdoctoral scientists, so if you are interested in working with me, please send me an email at ole2001@med.cornell.edu

Drug testing using zebrafish July 16, 2008

The Scientist has a short but quite interesting article on how some scientists are using zebrafish to test drugs. As the article says, the zebrafish is the only model system that has organ systems comparable to ours (it’s a vertebrate, it has a heart, blood, nervous system, etc), that also fits into 96-well plates … combine that with efficient techniques for creating transgenic animals (a company sells a library of 11,000 mutant lines), full transparency, and you get a system with an amazing potential for discovering potential drug candidates, predicting side-effects, understanding mechanisms of action, etc. Most likely, all that’s missing right now is a smart system for automated image analysis …

Regulatory elements in the coding region of odorant genes March 13, 2008

Odorant receptors (ORs) constitute one of the largest gene families in mammalian genomes. Odorant receptor genes are expressed in olfactory sensory neurons (OSN), with each OSN stably expressing a single OR protein (from a single allele). How exactly this is achieved is largely unresolved (and fascinating). How do cells chose which OR is going to be expressed ? How do they make sure that only this OR is expressed (which may means that other ORs have to be repressed). In a recent Cell paper (pointed out to me by Yoav), Ryba, Belluscio and colleagues present new exciting data about the underlying mechanisms. In mouse, transgenic odorant receptor genes cannot be expressed in olfactory neurons, when directly under the control of OR promoters. The authors hypothesize that this is due to interactions between regulatory sequences in OR coding regions, and regulatory sequences in OR promoters. So they tried to eliminate this interaction by separating OR promoters from coding sequences. They placed an OR coding sequence under the control of an artificial promoter (TetO), activable by TTA. The TTA-coding gene is itself placed under the control of an OR promoter. Using this system, they managed to express the OR in a good fraction of neurons. In turn, this allowed them to observe that the expressed OR suppresses expression of endogenous ORs. Likewise, TetO-OR expression appears suppressed in neurons where an endogenous OR is expressed. The authors conclude that regulatory sequences in OR coding regions are mediating the inhibition. Identifying these sequences and what binds to them is therefore the next step … it is unclear whether sequence conservation can help, as it appears that the ~1kb OR coding sequences are highly variable.

Determining nucleosome positions using Solexa sequencing March 9, 2008

In this amazing Cell paper, Keji Zhao and his colleagues used Solexa sequencing to create genome-wide maps of nucleosome positions in resting and activated human T cells. They digested DNA with micrococcal nuclease, then gel-purified ~150bp fragments (DNA fragments wrapped around nucleosomes are ~147bp-long) and ran their sample through the Solexa pipeline. The Solexa technology allows the 5′ and 3′ ends (~25bp) of these fragments to be sequenced, and these reads are then mapped onto the human genome.

The resolution of the results they got provides some fascinating insights into chromatin structure and regulation, and the link with gene expression. They detected 8 phased nucleosomes surrounding the TSS of expressed genes, but only right upstream (+1) of the TSS of non-expressed genes. The 5′ end of +1 nucleosomes peaked at +40bp for expressed genes, but +10bp for non-expressed genes. Nucleosome levels at -1 were lower than -2 and +1 for both expressed and non-expressed genes, suggesting that all core promoters are nucleosome-depleted. -1 levels of induced genes do not appear to change upon T cell activation, however increases in -1 levels were observed for repressed genes (-1 nucleosome levels appear inversely correlated to RNA Pol occupancy, also measured by ChIP-seq). The lists goes on and on.

While the authors focus their paper on nucleosome patterns surrounding the TSS, to me, nucleosome patterns in the rest of the genome would be more intriguing. What kind of nucleosome patterns are found in gene deserts ? in recombination hotspots ? in ultra-conserved sequenced ? they present anecdotal evidence that enhancer regions need to be nucleosome-free, in order for regulatory elements to be able to recruit transcription factors. It would be interesting to see if this is a general feature of enhancers (which would then allow to predict where they are in the genome).

23andMe December 9, 2007

According to The Economist, at least 3 companies (23andMe, deCODE, Navigenics) offer a new type of service where they will genotype several hundred thousand common SNPs from a DNA sample (e.g. saliva) you’ll provide. All this for a few thousand dollars or even less than $1,000, depending on the company. This is an interesting development, especially for us, scientists: if it catches on, the prices for this kind of services and associated technology will go down, and this will make it significantly cheaper to conduct whole-genome association studies with large cohorts of patients. This may also bring down sequencing costs even further and favor innovation in sequencing technology.

I have to say, I am tempted to try out one of these services. There are no genetic diseases running in my family, so I have no reason to be particularly afraid of discovering something bad (yeah yeah spontaneous mutations). I am just very curious to get a sense of what my genome looks like, and of how it differs from other genomes (from the HapMap people for example). Of course, SNPs are particularly interesting when they can be correlated to some phenotype, that is, if they can explain why we look the way we look, or predict disease risks. Provided customers of these services will be willing to give out enough details about themselves (with minimum self-reporting, and maximum objective tests), the companies offering these services could end up making significant discoveries in terms of genotype-phenotype associations.

Would I make my SNPs publicly available ? probably not. This is America, and like the article says, it does not sound inconceivable that health insurance companies will one day deny you coverage because you have an allele associated with increased risk for a certain disease. If doing a little research on the web can save them several hundred thousand dollars, it would be crazy to think they wouldn’t do it. Of course, I can always move back to a country that has universal health insurance if this happens

Nature Reviews Genetics highlight article about FIRE November 16, 2007

The article is nicely titled “Decoding the regulatory genome”.

http://www.nature.com/nrg/journal/v8/n12/full/nrg2281.html

The FIRE paper is out October 25, 2007

Molecular Cell table of content (Oct 26th issue):
http://www.molecule.org/content/current

Direct link to the paper:
http://www.molecule.org/content/article/fulltext?uid=PIIS1097276507006661

FIRE web site (to access the Web interface or download the software) :
http://tavazoielab.princeton.edu/FIRE/

The first two (mostly experimental) papers that used and cite FIRE have also been submitted this week, one to PNAS, the other to Nature Genetics.

Update Oct 29th: Nature Reviews Genetics is going to publish a “Research Highlight” on our paper.

Update Nov 8th: The paper was rated as ‘Must Read’ by the Faculty of 1000 Biology (F1000 factor 6.0)

Update Nov 10th: FIRE is also available for commercial licensing through Princeton University (totally free for academia of course).