User login
Navigation
Bezier curves crack me up
I just recently came back from a pretty disappointing systems biology meeting. I had fun and had some interesting conversations and met great people, but the scientific content of the submitted talks was really very variable. So variable in fact that I'll blog later about the concerns that this meeting raised in my mind about systems biology. I want to start off with something a bit more outrageous though: a wtf?!-style slide that was shown during one of the talks there.
The explanation given for this graph went along the lines of "... The expression dynamics of the system are clearly very complex ..." What? Are you joking? This graph is obviously nothing more than an example of the 8th deadly sin of overfitting.
Question: Can comeone over there in the Interweb come up with a valid use case for Bezier curve fitting in biology? I bet you can't.
- Matias's blog
- Login or register to post comments
hmm..
Firstly - what is a Bezier curve, how is it useful, and why do they use it? I've never been in close contact with them :)
Secondly, are you sure it's overfitting? The predicted concentration seems to be too simple of a model for the real data - I'd say the model is not powerful enough based on this plot alone..
So that good of a meeting, huh?
I imagine they used the curve
I imagine they used the curve fitting because it's available in Excel and looks nice and smooth :) There were no error bars or indeed any information of how many numbers of replicates each of those points (of noise?) correspond to. So, I wouldn't really even comment anything about the power or lack of power of their model from looking at that plot.
not a nice fit
if they were trying to fit that sinusoidal=like looking data, they have done a miserable job indeed :) You're right - something's fishy here.. Do you have the full talk?
strange graphs
Well, the two curves really don't fit....
Strange days...