I'm choosing not to talk about the abstract and data summary assignment until it's completely over, to prevent myself from putting my foot in my mouth when the first draft comes back and I've done everything wrong (well, hopefully not everything!). I decided to focus more on the scientific content of Mod 1 and I'll follow up about the writing portion later.
There are two ways to read this blog post. I'd recommend reading the non-ranty version first.
The first 20.109 module dealt with characterizing the gut microbiome in seagull populations from two different beaches, using sequence analysis. On the surface, this seemed pretty interesting and familiar.
My first UROP project at MIT involved some related material, looking at data from a human gut microbiome study and evaluating various community distance and similarity metrics. UniFrac, the phylogenetic distance measure we used in 20.109, was notably not one of the candidate distance metrics, for important theoretical reasons that weren't relevant in 109. The paper made it to publication, but I wasn't involved in it; my UROP mentor took this project as an opportunity to teach me a bunch of things while I essentially reanalyzed the data that had already been submitted for publication in search of more interesting conclusions. Rather than find something novel, I instead was only able to confirm the main conclusions of the study without adding much.
<rant>
[I'll reserve judgement as to whether that's a good idea in undergraduate research. On one hand, that project took a summer and didn't result in much fundamentally new science. On the other hand, the project was entirely my own, and while the question looked promising in the beginning the situation turned out to be exactly what we expected. C'est la vie--sometimes there isn't something interesting where you think it might be. On the (other) other hand, I left it with what is perhaps a much better grasp of ecological theory than I might otherwise have. I realize now that this probably served me very well during this module, and will hopefully serve me well in life.]
</rant>
Unfortunately, despite my high hopes for the module, snow days and the inherent time crunch associated with an academic class took a lot of potential value out of this module. While it served as a good introduction to molecular cloning skills and other wet lab techniques like primer design, the module didn't treat issues like ecological community analysis with a satisfying degree of depth.
[Disclaimer: what follows might read like a list of complaints, but I tried my best to justify everything]
<rant>
UniFrac was the only distance measure described in class, and while it's not a terrible measure, other distance metrics like Simpson's Index, Shannon's Index, and generalizations thereof were never touched. I mention this because using UniFrac over, for example, the Simpson Index (which, for some subfields of ecology, is the index of choice--UniFrac is by no means a clear winner) is a methodological choice. Given this data, perhaps it's the best choice, but it's worth mentioning why. Making justified choices about your experimental or analytical methodology is an important part of research, and 20.109 is as good a time as any to talk about it.
Principal Coordinates Analysis and Principal Component Analysis were used in software without really emphasizing what they mean. Importantly, the difference between PCoA and PCA is huge, and significant (PCoA deals exclusively in pairwise distances, while PCA deals with inherently multidimensional data--the kinds of things you can do with the output are very different; for example, PCoA cannot be used to make predictions about a new sample, while PCA can.)
</rant>
I'd like to emphasize again that the goal here wasn't to bash the first 20.109 module (as if I'm even remotely qualified to do so). Perhaps my "complaints" are nitpicks, and perhaps my hopes here were set too high. To some extent, I was hoping that going through Module 1 would make students more capable of working in the Runstadtler group, for example, than the average course 20 student, all else equal. That's likely not the goal of the module.
I imagine that executing a laboratory module that deals with current research but must also introduce students to fundamental techniques is a hard line to walk. The broader scientific scope of this module (which is fundamentally an ecological one) was quite far removed from what most students in course 20 will have been exposed to, even after mostly completing their degree, so that makes the task even harder. Given how hard it is to teach this module, I think this was very well run. However, I might question the inclusion of this module over an alternative module in future incarnations of the class, if only to make the job of balancing extraneous theory with laboratory fundamentals a little easier for everyone involved.
There are two ways to read this blog post. I'd recommend reading the non-ranty version first.
The first 20.109 module dealt with characterizing the gut microbiome in seagull populations from two different beaches, using sequence analysis. On the surface, this seemed pretty interesting and familiar.
My first UROP project at MIT involved some related material, looking at data from a human gut microbiome study and evaluating various community distance and similarity metrics. UniFrac, the phylogenetic distance measure we used in 20.109, was notably not one of the candidate distance metrics, for important theoretical reasons that weren't relevant in 109. The paper made it to publication, but I wasn't involved in it; my UROP mentor took this project as an opportunity to teach me a bunch of things while I essentially reanalyzed the data that had already been submitted for publication in search of more interesting conclusions. Rather than find something novel, I instead was only able to confirm the main conclusions of the study without adding much.
<rant>
[I'll reserve judgement as to whether that's a good idea in undergraduate research. On one hand, that project took a summer and didn't result in much fundamentally new science. On the other hand, the project was entirely my own, and while the question looked promising in the beginning the situation turned out to be exactly what we expected. C'est la vie--sometimes there isn't something interesting where you think it might be. On the (other) other hand, I left it with what is perhaps a much better grasp of ecological theory than I might otherwise have. I realize now that this probably served me very well during this module, and will hopefully serve me well in life.]
</rant>
Unfortunately, despite my high hopes for the module, snow days and the inherent time crunch associated with an academic class took a lot of potential value out of this module. While it served as a good introduction to molecular cloning skills and other wet lab techniques like primer design, the module didn't treat issues like ecological community analysis with a satisfying degree of depth.
[Disclaimer: what follows might read like a list of complaints, but I tried my best to justify everything]
<rant>
UniFrac was the only distance measure described in class, and while it's not a terrible measure, other distance metrics like Simpson's Index, Shannon's Index, and generalizations thereof were never touched. I mention this because using UniFrac over, for example, the Simpson Index (which, for some subfields of ecology, is the index of choice--UniFrac is by no means a clear winner) is a methodological choice. Given this data, perhaps it's the best choice, but it's worth mentioning why. Making justified choices about your experimental or analytical methodology is an important part of research, and 20.109 is as good a time as any to talk about it.
Principal Coordinates Analysis and Principal Component Analysis were used in software without really emphasizing what they mean. Importantly, the difference between PCoA and PCA is huge, and significant (PCoA deals exclusively in pairwise distances, while PCA deals with inherently multidimensional data--the kinds of things you can do with the output are very different; for example, PCoA cannot be used to make predictions about a new sample, while PCA can.)
</rant>
I'd like to emphasize again that the goal here wasn't to bash the first 20.109 module (as if I'm even remotely qualified to do so). Perhaps my "complaints" are nitpicks, and perhaps my hopes here were set too high. To some extent, I was hoping that going through Module 1 would make students more capable of working in the Runstadtler group, for example, than the average course 20 student, all else equal. That's likely not the goal of the module.
I imagine that executing a laboratory module that deals with current research but must also introduce students to fundamental techniques is a hard line to walk. The broader scientific scope of this module (which is fundamentally an ecological one) was quite far removed from what most students in course 20 will have been exposed to, even after mostly completing their degree, so that makes the task even harder. Given how hard it is to teach this module, I think this was very well run. However, I might question the inclusion of this module over an alternative module in future incarnations of the class, if only to make the job of balancing extraneous theory with laboratory fundamentals a little easier for everyone involved.
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.