Bioinformatics

[Radagast]

There was a good side-bar conversation going on tree methods in one thread. I thought it might be good to continue it here...

Do you know about the Neighbor Joining method? It's great in that you can actually do it by hand, and my professor found a tutorial online that allows you to walk through it by hand. http://www.icp.be/~opperd/private/neighbor.html is a really good tutorial, and I believe it uses the Jukes-Cantor substitution model if I remember right. It doesn't teach you about testing the node, like bootstrapping--one of the most basic. You could do bootstrapping by hand, but I wouldn't recommend it considering you'd probably do at least 500 bootstrap replicates per node. Just know how to do one, then use MEGA do all of the calculations for you.

The best way to learn phylogenetics is by just doing. Download MEGA and MrBayes, do the tutorials, and get the trees. I'm sure you've heard of GenBank by now. Develop your own miniproject by grabbing sequences off of GenBank and use the various tree building methods in MEGA and use MrBayes to build trees. See where they're similar and we're they're the same, look at the values of the nodes, when should you more comfortable with the nodes? Does a high value mean a correct node? In distance trees, what is long branch attraction?

Know what Maximum Likelihood, Bayesian Inference, and Bootstrapping are supposed to do, but not the math behind it. Know what Markov Chain Monte Carlo does, but not the math behind it. Know the various substitution models, and know the math on the ones that aren't so parameter heavy.

Most importantly, know that if you see a tree in the literature, realize that it may be wrong only because of poor alignment, and it may be something that is out of their control, and it may be an issue that can be solved by knowing a little more about the organisms and how they got to where they are and such.

Hope that helps.

 

[Blayz]

phylogeneticists aren't true bioinformaticians.

 

[Radagast]

No?

 

[Blayz]

It was a joke.

sheesh!

 

[ArnautDaniel]

Its all interesting stuff.

I've been reading up on attempts to apply the same sort of analysis to indo-european languages.

Of course that analysis is rather more iffy than with biology.

 

[Blayz]

Its all interesting stuff.

I've been reading up on attempts to apply the same sort of analysis to indo-european languages.

Of course that analysis is rather more iffy than with biology.

It's not that iffy. One of the more interesting phylogenetics talks I have seen involved the evolution of a chain letter (one of those "pass this on or bad things will happen to you" things).

However, lets move on, what do people thing of the challenges of the new deep sequencing technologies? Do you vote 454 or Solexa? Wil they kill chIp chip? Did microarrays ever live up to their hype?

 

[ArnautDaniel]

It's not that iffy. One of the more interesting phylogenetics talks I have seen involved the evolution of a chain letter (one of those "pass this on or bad things will happen to you" things).

However, lets move on, what do people thing of the challenges of the new deep sequencing technologies? Do you vote 454 or Solexa? Wil they kill chIp chip? Did microarrays ever live up to their hype?

Well by the time you've chosen your core vocabulary that is supposed to be resilient to change (but probably isn't as resilient as you like), and then you've selected the cognates (and hope you haven't mixed up a borrowing or some like, and haven't assigned false cognates as cognates) and so forth, it has become rather iffy.

It all depends on the quality of a data set that has already undergone a large amount of selection.

 

[TheGnome]

Haha, I honestly didn't know I was talking about bioinformatics. There's a new class being taught this semester on bioinformatics that a few of my friends are taking. I didn't really know what it entailed, and I was going to register for the course but opted out to take a seminar series instead. I'm glad I did, I asked my friends what the class was about, and they still don't have any idea. The exam kept getting pushed back, and they still haven't had their first exam even though it's mid-October already. Hell, I'll be taking my second exam in molecular microbiology on Tuesday, and I believe their first exam will be on Thursday.

Bioinformatics may probably be one of those useless broad categories like genomics. It's a really sexy name for a category, but it really explains nothing. Phylogeneticists are genomicists; population geneticists are genomicists; molecular geneticists are genomicists and proteomicists. Whatever happened to good ol' fashioned "geneticist"?

 

[Blayz]

Well by the time you've chosen your core vocabulary that is supposed to be resilient to change (but probably isn't as resilient as you like), and then you've selected the cognates (and hope you haven't mixed up a borrowing or some like, and haven't assigned false cognates as cognates) and so forth, it has become rather iffy.

It all depends on the quality of a data set that has already undergone a large amount of selection.

woooooooord saaaalaaaad

unsupervised clustering. Let the terms rank themselves. Text analytics is quite an advanced field.

 

[ArnautDaniel]

woooooooord saaaalaaaad

unsupervised clustering. Let the terms rank themselves. Text analytics is quite an advanced field.

Well the user still has to tell the computer which languages they (the user) think have cognates in which categories.