Saturday, May 9, 2015

A lot of biology work is slow.  XKCD made the joke about how course 6 people have nothing to do while they wait for their computers to work, but biology is another game entirely.

Now, we don't do too much chair jousting,

you try to do several things at once, but these things take time.  Take making a plasmid.  You have to wait the few hours for the thing to ligate and such, then you have to do a transformation, plate that, let the plate grow, pick a colony, run a miniprep, digest and gel it to verify the insert is there, then probably do sequencing.  That's a few hours of reactions and 2-3 overnight steps.

And that is if nothing breaks.  I learned in 7.03 that a lot of statistics for science came out of biology because other sciences just give you answers, biology can at most give you a strong probably.  My teams mod 1 gel failed perfectly, neither of our DNA lanes had product but our DNA- control had a nice bright band of contamination.  The whole class had some great news when the whole batch of sequencing reactions came back as having all failed.  These things happen all the time in biology, living things are messy.

Some of this stuff is also really poorly optimized.  I think that its feasible with today's technology to never require another human hand to do a mini prep.  Most of the lab work could probably be improved mechanically as well.  The protocols could also probably be improved.  For example, the microbiome study we did could use some major streamlining.  Our process was to use a prep kit to purify the bacterial dna, blunt end clone it, transform it, pick colonies, miniprep those, then sequence them.  This could be trimmed to a tiny fraction of this.  Almost no work is needed to prep the stool samples for pcr, you can autoclave the whole batch of samples to disable most of the PCR blocking proteins then dilute it until the rest are too few and far between to have an effect.  The diluting strategy works because PCR is really powerful.  Next, you can use PCR primers that stick BsaI cut sites on the ends of the DNA and use restriction ligation reactions which are much cheaper and more robust to make plasmids.  Finally, you can basically just mail the plate from the transformation out and have as many colonies as you want sequenced.  True story, you can send Genewiz an agar dish with colonies and two tubes of primer (forward and reverse), and they'll send you back the sequence.  That's a few hours to autoclave, PCR, clone, and transform, then an overnight to grow and then its out the door.  All of those steps are also just pipette into the tube then put it in the machine and wait steps.

At the same time, we managed to get a lot done.  In module 1, we had about 1000 bases of sequence, *2 bits/base, *100 reads, *1 byte/8bits, which works out to 25kb of data.  Module 2 actually produced some statistically valid data.  Module 3 would have probably done a lot if only things had worked a little better (I felt like a gold rush era prospector at the TEM).

Most of biology works like this.  We were told in class that a lot of biology veterans have tendonitis from all the minipreps.  Some of the early biology experiments scare me (and not just the unethical ones).  The plant hormone ethylene, for example, was discovered because it makes plants grow in strange shapes.  Having to use that as an assay for activity instead of GFP must have been unpleasant.  At the same time, we have a lot done so far.  We basically figured out that biology was made out of chemistry at the end of the 1800s.  We figured out the structure of DNA in 1953, and we finished sequencing the human genome in 2003 (50 years, that beats the Wright brothers-moon landing gap by about a decade).  Now we have the cost of sequencing down in the $1000 range.  I think XKCD sums up what biology has done so far, and they're talking about infectious disease, we fixed that with vaccines and antibitoics way back in the day.

Unfortunately, it seems like we can't quite do all that much in a one semester lab class.  Biology takes a lot of time we don't have.  Perhaps one day we can produce publishable work, but we aren't there quite yet.  The biggest problem we had was small sample size and time constraints.  I'm pretty sure the p-test we did in mod 1 said we didn't have enough data to say anything.  Module 2 said that one cut type had a difference in effect, but I think we could have ordered the other ones if we had more data.  Module 3 was a wash because we didn't have much gold in our samples, which is what we were trying to test.  We also did some strange things in lab, like start an overnight on Thursday and finish it on Tuesday.  The project proposals might be useful though.  I know mine is completely doable, as soon as I don't have a conflict of interest I'm going to submit it to my UROP lab.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.