Siliceous and Delicious

The day of Heroen Verbruggen coming to visit is here. It’s been a total waste—since I’d been hoping to talk to him about the morphospace, but haven’t had a chance to work on it because of the microscope fiasco. Fucking shitballs. I had breakfast with him at Darwin’s and we chatted casually about this and that, but nothing of substance or really relevant to my work—he did ask about the morphospace project, but I had to answer honestly that it was in the same place as it had been when I’d last emailed with him. That was a quick way to kill that conversation. I can truly say that, right now, I hate science, and the mere talk of it makes a sickening feeling spread through my interior.

Eventually I ran out of things to talk about and things to show him around campus, so I brought him back to the lab at around eleven and sat him down in the library to get on with his own work. Again, I feel like a douchebag for having asked him for help and then not having anything to really talk with him about—but what can you do. It’s just another indication that these things are not meant to work out for me.

“Did” a few more slide species checks, in which it became increasingly clear that I need help in identifying these species. For the first few species I knew pretty well what I was looking for, but for the last two in particular the species descriptions in the taxonomic literature just weren’t clear enough for me. This is where I would want my advisor to sit down next to me and help me, but of course, Andy isn’t a radiolarian specialist, and would be just as lost. So, what next? Emailing pictures to Dave, asking him whether I’ve got the IDs right? Traveling to Berlin? It certainly dealt yet another blow to my confidence.

The rest of the day was pretty well wasted away with various Heroen babysitting activities, sitting through his rather long (90 minute) talk, going to coffee with him, and now going to dinner with him and Erik and Andy. It should be fun, yay, an opportunity to network and a free meal—but I am so sick of the whole scientific endeavor at this point that I can barely imagine anything worse at this point than being forced to sit around a dinner table and talk enthusiastically about science and how great it is for two hours.

On a slightly positive note, did a few minutes worth of googling about digital educational publishing at one point, daydreaming about fulfilling work in a startup company, and found this rather neat looking outfit. Of course they are based on the wrong coast, but nice to think there are some fun places one could conceivably work after this torture is over.

Shitty day. No word from committee about possible meeting dates. Plugged away at microscope today, all day long, checking to see if species are there. Haven’t tabulated the results, just blindly scribbling down notes, as to not discourage myself completely. It doesn’t seem to be going well, though—many of the slides don’t have any of the species I need. Many others are confusing—I don’t understand the taxonomy and will need more schooling to be able to ID them.

All in all, things are going fairly poorly. Motivation low. It’s raining. Shitty day.

After yesterday’s near-meltdown, thankfully prevented by some DSA-administered seawater to keep the angry fuel rod casings from releasing rageioactivity, back to grappling with the imaging issue today. Checked online and found a PDF describing the Olympus lenses I have to play with—the nominal working distance for both of them is around half a millimeter. Measuring the thicker slide I had trouble with yesterday confirms that the sample is about 1mm from either side of the slide—beneath about 1mm of canada balsam plus cover slip on one side, and about 1mm of glass slide on the other. So no hope really for those thickness slides with the objectives I’ve got, plus any that have more than about .4 mm of canada balsam, which is many of them (I’m guessing half or so).

Trawled the internet a bit more to find out about long working distance objectives. It would appear that something like the Olympus LWD CD Plan would do the trick—it has a nominal working distance of 2.04 mm, which ought to be enough for even the most heavily embedded ones. It only has a numerical aperture of about 0.55 though, so the questions then are:

1. How much does an objective like this cost?, or
2. Does Jacques have one I could borrow, or know someone who does?, and most importantly, if so
3. Are the thinner-walled shell thicknesses measurable with a 40x objective
4. Do any of the slides actually have the species I wanted on them?

I should be able to answer number 3) by using a properly prepared, non-canada-balsam-ocean slide with thin-walled rads on it, and seeing how thin of a feature I’m able to resolve using the camera system. But I also need to, of course, still answer the overarching question, number 4).

First, I did a bit more blind googling and found another, better Olympus lens—the LUC Plan FLN 60x, with a numerical aperture of 0.7 and a working distance of 1.5-2.2mm, plenty to get right in there. The main drawback I can see—besides not knowing whether it would work for imaging stuff under a coverslip—is that it costs north of $2000. Then again, this might be what makes my thesis a reality. However, as I am finding out on further reading, these new Olympus lenses seem to be made for infinity-corrected microscopes, rather than the fixed tube length (of 170 mm) design of our microscope. Pants.

Anyway. I wasted a good couple of hours googling around for lenses, and all I could find in the end were indications in old catalogues that suitable lenses (like the first one mentioned) exist—but none for sale anywhere. The closest I could find was here, a 60x Nikon LWD objective, although it doesn’t say how long the working distance is on it and I couldn’t find the appropriate Nikon catalogue to tell me. So much for question 1). I did also walk over to Jacques’ office to see if he would sign my progress report form, and see if he had any helpful objectives, or perspectives, or even adjectives, but he wasn’t there. So question 2) is not answerable right now.

Well, I’m pretty well shit out of ideas on this front so I’m moving on to answer question 4). On to the microscope.

One more slide had nothing… and one had at least six individuals of one species of one of my lineages! Hallelujah! A small success, a small relief, at last. There is hope. Somewhere, there is hope!

Been working on calibrating the microscope so that I can actually determine how good (or bad) the resolution is of the objectives that are available and seem to work with the ODP slides. Andy was finally back today, so I was able to get at his stage micrometer and begin that laborious process. Set the pixel scale for the 4x Olympus objective, then jumped right up to the 40x Olympus SPlan, which I found in a drawer in the microscope room, which has a numerical aperture of 0.70 and thus should theoretically perform better than the 40x Olympus DPlan PO that Andy gave me—not sure if 0.05 makes a big difference in numerical aperture, but I do know that bigger is better. The aperture on the 63x oil immersion objective I had been hoping to use (just for sake of comparison) is a whopping 1.40. Never going to get anything near that good working considering the lenses on offer here, and the irritating way these samples seem to be prepared. Ah well.

Once I had calibrated that 40x objective, I figured I’d measure some shell thicknesses and see how thin the shells could get before I lost resolution. I measured a shell thickness of 2.8 µm without much trouble. Nice, but at the very thick end of what I expect to see for shell thicknesses—the slide was also one from the Oligocene, Pleistocene shells will be much thinner (depending on species, of course). Anyway, I thought I’d try a Pleistocene slide to see what I could see, when the next level of total disaster struck—this slide was so thick I wasn’t even able to use the 40x objective. Fuck. Fuck a ducky duck. Even though this objective has a larger working distance, it’s not enough to image through the bottom of the slide (i.e. the main slide glass)—so, this fucking sucks.

What I realized yesterday, after a completely unproductive week, is that I’m not ready to ignore what’s going on with the radiolarian lineage project and just focus on the morphospace. The stalled radiolarian project is making me nervous, worried, and scared of failure. That’s no place to be emotionally. I need to tackle it head-on, and try to figure out how I can make that project work. It needs to work, because I need to graduate, and I’m going to make it work. So what if Andy isn’t going to help me. He’s not going to do my PhD for me anyway, I’m going to. Fuck it.

What’s going to make me feel better is knowing, for sure, just how good the resolution I can get from the functional objectives is, and if I know how many of the slides actually contain the species I’m looking for. Both of these questions are not hard to answer.

The first requires a stage micrometer—a microscope slide with micron-scale markings on it that allow calibration of the microscope objectives—an object we have in the lab, but hidden away in one of Andy’s desk drawers. Behind a locked door. And Andy is away. It’s a bummer but I’m considering using my key to his office, handed down through the generations to the senior Knoll lab student, for precisely such emergency situations. Of course, I don’t want to get caught red-handed having let myself into his office. But it’s a risk I might be willing to take considering the pressing circumstances.

The second is also easily done—sit down at the microscope with the species description and the list of slides, and see what I can find in each. This, in fact, is something I can do right away. Got through one slide (trekked to the Office of Career Services for a much less interesting talk than the one yesterday over lunch), which didn’t have any of the target species in it. Oh well. Better luck with the next one. Also spent a good hour or so researching the offending objective online to see if I could find out what it’s nominal working distance is—and in fact what the working distance means. The former was eventually found (though it required a tremendous amount of sleuthing)—the Leitz PL APO 63/1.40 objective has a nominal working distance of 0.15 mm, i.e. 150 microns. It is corrected for a cover slip thickness of 0.17 mm. This means, to the best of my understanding, it will focus on a plane that is 0.15+0.17 = 0.32 mm, or 320 µm, beneath the tip of the objective. That’s not a long distance. It is, however, apparently also a very fine objective (for its designated purpose), retailing in the second hand market for close to $4,000 as far as I can tell.

Anyway. The next tasks are, as I decided earlier, to see if the species I need are actually there, and if the objectives I do have—specifically the 40x Olympus—offer enough resolution to measure shell thickness on them.

Managed to faff away the first half of the day—I think the disappointment of finding out yesterday in an email from Dave that high-powered objectives might just be out of the question for the radiolarian project (potentially—just maybe—leaving the project mortally wounded) finally sank in and manifested itself in a general baseline malaise and grouchiness this morning. Eventually settled down to being coding characters for the diatom morphospace matrix, so many years coming, after lunch. Halle-fucking-lujah!

One question about this process, which I’ve kept in the back of my mind, but haven’t answered so far, is how to code for missing and inapplicable characters. Kevin Boyce’s matrices have only one character state for all eventualities, a question mark. I suppose at the analysis stage it all comes out in the wash, and it doesn’t matter whether a character is unobserved or inapplicable, it just won’t count towards calculating morphological distance between taxa. However, there does to me seem to be a fair bit of information in making that distinction—a character that’s theoretically knowable but just not clear in the pictures, descriptions etc available versus a character that describes something not present in the taxon. I plan to differentiate these by using “?” for characters that are unclear or unobserved and using “n/a” for characters not applicable to the taxon in question. This can swiftly be turned into all “?”s in the data analysis step, but going back the other way wouldn’t be easy!

A related but slightly different issue is how to code variability—when a genus displays more than one character states for a given character. Kevin dealt with this by making an extra character state in such taxa, called “variable”. I’ve tried to avoid this issue in my choice of characters, but I’ll make the judgement call as I go through the coding as to whether I feel I need to add a “variable” state. In part this might be problematic because for many taxa I won’t have a huge range of images or descriptions easily available; there’s also the problem of how to calculate distances between the “variable” state and other states—are they less than the other inter-state distances, or the same…? Better to avoid the problem, if at all possible… We’ll see.

Spent the morning following up with the next step on the radiolarian project: since we couldn’t get the microscope working, and thus I couldn’t test the measurement protocol, I had to email Dave for help. This involved writing an update on my research (which ended up being rather lengthy, but was something I would have needed to do sooner or later anyway) and chewed up the whole morning, but it was what needed to happen.

Not sure how to follow from here. Fuck around with the rad project until the measurement set-up is perfect? Get demotivated because it’s been weeks, and the microscope still doesn’t work? Because Diane didn’t order the part and it got delayed and…? It’s my worst-case-scenario brain kicking in, but I can’t help it, and the best cure I can think of is to let it lie for the moment—at least until I hear back from Dave about suggestions for what to do about the objectives—and focus on the morphospace. Get that baby done and feel good about life.

An unsatisfying day. At least I feel justified in having started late and ending early, since it was so frustrating. Spent most of the morning trying to get the damn camera to cooperate with microscope and Canon’s own software. Much of that turned out to be my own idiocy, since I was baffled as to why the software wasn’t allowing me to control the camera’s aperture. This lasted for a long time, until I realized (sound of penny dropping) that since the body was mounted to the phototube without a lens attached, there was no aperture assembly to control, hence no setting to make. Duh. Eventually I resolved the problem I was having with the live preview being much darker than the captured final image (setting with shutter speed priority at 1/125 seemed to mitigate the issue, goodness knows why). In the afternoon, however, the next frustration, this one less tractable—couldn’t get an image out of the 63x oil objective (the highest power on the scope). No matter how hard I tried, no image. Eventually caved and asked Andy for help, who at first demonstrated knowingly various condenser settings I hadn’t considered, but after futzing and trying for a good half an hour, and trying a couple of other lenses on his microscope and from his drawer, came to the conclusion that he couldn’t get an image either. He apologized, but didn’t really have any good suggestions for how to go on. He seemed to hint at some point that the 40x objective was also really good, but I made it absolutely clear that that wasn’t going to fly—I was using a 100x objective in Bristol and Berlin to make the shell thickness measurement, and I wasn’t going to settle for something that much crappier (it also just won’t work, full stop).

I feebly suggested that Dave might have an idea of what’s going on. The best suggestion I got out of him was that, since insufficient working distance might be the problem, there might be a better objective out there for the job, and if so, that we could buy it. I’ll send Dave an email tomorrow—I promised him a report of  how my research is going anyhow—but today I’m done and heading out for wedding planning chores galore.

Had a tricky decision to make this morning—get the interface finished and done, or tackle one (even just one!) of the list of deliverables I committed to for DSA last week? Well, it’s been a very busy week with non-thesis commitments, so I decided to put off the non-thesis deliverables for just a little bit longer, and give priority to the one thing that really needs getting done more than anything else.

Knocked the first item off the list before 10am, which was to sort out how the magnification/scale information was going to be saved to the filename. I wanted a way to make absolutely sure that the scale was being set for each image that was opened . In almost every case there will be multiple images per individual, and quite possibly multiple images between write-to-pipe events, so it wasn’t sufficient to use those file-write events as a trigger for a filename change. I thus had to dig around the interwebs quite a lot for a way to get at the somewhat lower-level file opening event as a trigger for running my script. Eventually found a java class “Image Listener” that will do just that, and after a bit of tinkering I got it to both work perfectly, and also to compile automagically upon startup (to avoid having to install the plugin every time ImageJ is opened).

Next on the list was to write a function for the R interface that writes a backup of the SQL database tables to CSV files, to avoid any disastrous things from happening should the .db file become corrupted or some such nasty yet unforeseeable thing. This was not difficult to accomplish and was done by lunchtime.

Next, I tried to improve the “New slide…” menu item to allow for accidentally adding a slide that already exists, and for choosing an existing slide to work on. Completed this task after DSA, and it seems to work fine. I’m ready for testing!

Put together a report for Andy late last night for our meeting today. It was an interesting exercise—made me realize a) just how much I’d actually done in the last three months, b) what a good call it was, motivation-wise, to stop working on the outline morphospace Sébastien when I did, before things spiraled out of control, and c) how incredibly useful it is to keep this journal of what I’m doing. There’s no way I’d have been able to give a full accounting of what I’d done since November without the help of this blog; in fact, when I sat down to start writing, I had no clue what I would write—I couldn’t remember really having done much at all. How wrong I was.

Before our 10:30 meeting (which went on for almost an hour and a half), I quickly set up my laptop and rushed to fix all the little things (application settings, macro installations, working directories, path names for file references) that would allow the RadData interface to run in the microscope room so I could demo it to Andy. Got it working just in time, though I realized in the process that there’s another couple of features I need to add to the interface. Inserting a slide with the “New slide” menu option is required before the first measurements are recorded in order to set the current_slide variable, but the resulting SQL INSERT fails (as it should) if the slide entered already exists. Two things thus need to happen:

1. The exception thrown by inserting an already-existing slide to the slides table needs to be handled so the program doesn’t crash in that case.
2. The user needs to be presented with an additional menu option in order to choose a slide already existing in the database.

Andy’s feedback was mostly of the usual kind—looks great, I agree with you, do what you say you’re going to do, you’re on the right track, etc. He did have a fair few things to say about the mathematical morphospace, most of which I found elicited a “you really have no idea” type of frustration. He was quite keen that I continue on with that line of inquiry—eventually, at least, since he agreed that it was smarter to focus my energies on completing the morphospace first—and “not make it too complex”. He pointed out that the genius of Dave Raup’s groundbreaking work on the morphospace of coiled shells was that he was able to think about it simply, identify the smallest number of parameters that were likely to be functionally consequential, and ignore the rest of the complications. This, of course, is easily said, and is exactly what I set out to do initially when I first started thinking about this problem (over three years ago!), but I’ve come to realize that it’s naïve to think that this sort of solution is possible for all sets of biological shapes, just because it’s possible for coiled shells. I wouldn’t for a second suggest that Raup is anything short 0f utterly brilliant, but he also chose the right problem to work on. It’s just the case that coiled shells can almost all—in a general and imprecise way—be represented by a simple three-parameter geometric model. It’s my belief after wrestling with diatoms for the last three years that the same is just not possible for diatoms. There’s an underlying developmental process in how coiled shells are produced biologically, regardless of whether they’re clams or snails, and that process is easily captured by a model that describes a generating curve rotating through space. The developmental processes that lead to diatom frustules are fundamentally different, for starters occurring at the cellular level, mediated by the cytoskeleton, which moves cell compartments into a much harder to define array of different structures.

I voiced my concerns about the potential functional meaning of looking at just one cross-section of a complex 3D structure (i.e. the outline shape), but Andy dismissed my concerns with the argument that most of the shapes are really just flat-topped or domed, and that “the silly shapes really don’t make up much of the morphospace”. Well, I just don’t know if that’s true. Sure, many of the most abundant forms are simple in their topography, but isn’t the point of a morphospace to look at the realization of successful morphologies in a broader space of shape possibilities? Anyway. I smiled and nodded and thought to myself, “well, let’s see if I get to it, but I think I’ve tried this, and I’m probably going to leave it up to someone more genius than me to solve this problem”.

The crowning glory of the meeting was a demonstration of the RadData interface, in its first sorta-kinda working form. I was astounded by how little Andy understood of the system as I demoed it… There was a giant chasm of understanding, actually, and it made me both very proud and a little bit sad. He was clearly quite impressed, but also didn’t really know what to make of it, I think. He was nodding attentively as I explained how it worked, and I thought he was on board, but as I was demoing the live preview on the Canon camera control app he asked whether the computer knew which of the specimens on the screen it was supposed to measure, I realized just how much more basic my explanation would have to be to get through… But that’s way more than could be fit into a one and a half hour meeting.