AI

I think the scary part is how the robot hand went from ponderously and kind of clumsily hardly being able to do anything useful a generation ago - skipped straight over being equally as good as a human hand - to being faster, more accurate and more dextrous than a human hand in this video.  Either it can't bounce a ball at all or it can bounce a ball a thousand times a second and you can't even follow it with your puny human eyes.

It's not AI - but you can suddenly see what they are on about with the idea of a singularity - in the same week that AI is able to operate as intelligently as a human mind - they will be 100 times smarter than us.

GWAS

Any gene (allele) that caused disease by itself would be strongly selected against.  A more realistic scenario is an allele that causes a disease when paired with some independent alleles and prevents disease when paired with other independent alleles.  This would not be discovered with current study designs.  An allele that caused disease only when paired with some other alleles (never prevented disease) wouldn't even be found.  The fact that this is not mentioned in this story and not mentioned in much of the gene/disease association study literature makes me angry.  I feel like everyone is stupid or hiding the truth on purpose.

biology ramblings

One of the major limitations of current biology research is the availability of sufficient quality and quantities of biological material.  Cell culture systems are available that recapitulate many important biological processes, but in many ways research is limited by the limitations of these systems.  In the future several avenues could be used to overcome current limitations in the biological material for research.  The first is the creation of new cell culture systems and the development of new method to collect primary tissue samples.  Clearly, this type of advance will give researchers new systems and source of material to study.  The second avenue is to improve genomic assays to work on smaller quantities of starting material.  This would allow work on highly pure specimens of rare primary cells types that cannot be isolated in sufficient quantities for current methods.  Ideally, it would be possible to assay the genome-sequence, histone modification pattern, DNA methylation state, transcription factor occupancy, and RNA content of a single cell.  

Genome Wide Association Studies

Maybe I'm misunderstanding something, but as far I understand it, genome wide association studies look for single polymorphisms that are correlated with disease, and do not currently have much power to detect combinations of polymorphisms.

I predict that combinations of polymorphisms will explain a large part of the link between genotype and health.  New algorithms to do this mike incorporate additional information linking genes and loci to increase power.

Podcasts worth listening to

Here is a list of the podcasts that I listen to.

Savage Love
This American Life
The Moth Podcast
WNYC's Radiolab
Common Sense with Dan Carlin
Dan Carlin's Hardcore History
The New Yorker Fiction Podcast

ChIP-Seq analysis

In ChIP-Seq data analysis the signal behaves nicely when transformed into the ???? space.

A logarithmic transformation does not behave as nicely.

White Squirrel

I saw a white squirrel in my backyard in Somerville, MA. It was eating a piece of toast with jam.

Random Process?

If you have a complex progess comprised of some deteministic processes and a random process, is the complexes process random or deteministic?  What if there are multiple random proecesses in the overall process?

By 'deterministic process' I mean a proces whose current state can be predicted for it's original state.

Roomba

My wife bought one of these a few days ago.  I thought it would be shitty, but it is not, it works very well.  If it keeps working this well, and lasts for two years, I'll think it is well worth the $300 we paid.

A method for analyzing the ratings in a music library and suggesting new songs

There should be a plug-in for a media player (I use MediaMonkey) that analyzes you music library and