I'm now in the main hall for the final keynote presentation. The hall is packed with people, currently there is a promotional video for Brazil, which is where ISMB2006 will be held. I'm definitely requesting travel funds for that, but then again, I'll have to travel with the boss... We have the governer of Michigan here apparently, Jennifer Granholm, will appere ? She's here, she talked, we discovered that Michigan is shaped like a hand.
Pavel Pevzner's key note has just started. He's talking on the topic of genome rearrangements
- Two types of rearrangements
- Small and large
- Are there similarity blocks in the genome ? how do we find them ?
- What is the architecture of the ancestral genome
- Genome consists of 280 architectural blocks or synteny blocks
- What are the evolutionary forces that effect genome architecture
- Whole genome duplication
- There is recent evidence for this
- Random breakage model (RBM)
- Was reiterated in hundreds of papers
The examples that he is using mainly concern X chromosome in mouse (and human ?). He discusses rearrangements in evolutionary context for example in tumor genomes, rearrangements may disrupt genes and alter gene regulation, translocation in leukemia "Philadelphia" chromosome. Cytogenetic analysis of the breast cancer tumor genome
suggests complex architecture, what is the series of rearrangements that leads to this state ?
Definition of a Reversal - kind of rearrangement. Take a series of 10 genes - what is the shortest series of reversals to transform one genome to the other. Looks like some kind of dynamic programming technique to find the shortest path through a series of reversals i.e. change one genome to another...
Duality theorem - Break point graph:
break point distance = #elements + 1 - #cycles + #hurdles
Multi-chromosomal rearrangements are more difficult to model e.g. translocations. Something about concatenating two chromosomes to devolve a translocation into a combination of reversals. This means we can use existing theory to analyze more complex translocations. Pavel claims it doesn't work. He mentions that there is a server available to do this rearrangement analysis: GRIMM server
He describes the GRIMM-Synteny algorithm and the general problem of synteny block generation, his paper on this is available here, read the paper I won't even try to explain it here (mainly becuase I don't think I appreciate the details). Describing how the system works... difficult to follow for me... he describes constructing a break point graph... HP theorem and GRIMM-Synteny reveal the evidence for rearrangement... ?
The big question Pavel is trying to answer is: are there rearrangement hot spots in the human genome ? This guy is basically building models for fundamental processes in human genome architecture. And they are doing it very thoroughly. It is one of the great things about these conferences to see some of the best people in their field discuss their work. Pavel has a pretty laconic presentation style too. Claims that due to proof not fitting theorem that the same break point is being used over and over again (explaining why some of the numbers from the model don't agree. At this point I'm having a little trouble following the logic...
95 % of the genome is in synteny blocks, 5% is around the break point regions. Mouse genome analysis suggests that break points may reoccur in certain regions: break point re-use and rearrangement hotspots. They suggest a new model: Fragile breakage model. Genomes are mosaics of fragile regions - about 5% and solid regions - the rest - with low propensity for rearrangements.
David Sankoff criticized the numbers from Pavel's paper, due to Due to bad parameter choice. Their alternative to GRIMM is ST-Synteny. Pavel is going through their algorithm pointing out the flaws.
The paper is here:
Chromosomal breakpoint re-use in the inference of genome sequence rearrangement (D. Sankoff & P. Trinh) Journal of Computational Biology, 2004, in press
Pavel claims that their algorithm is flawed they screwed up their synteny-block generation algorithm, even though they emphasized the importance of this part of the problem. They also never tested their algorithm on real data. And this is why his approach is better. Acknowledgments, crowd claps loudly, seems the tribe is happy with the performance of one of their stars...
Questions...
Too hard to follow... many sycophants... claims that looking at real data is significant... I run to the bar...
