ucam development 25 Feb 2010 from mac cowell on Vimeo.

I’ve been fascinated by bioluminescence from the first time I saw pictures of deep-sea angler fish. Since then, I’ve seen dozens of live bioluminescent creatures at the Boston aquarium and cultured E. coli expressing the LuxCDABE operon from Vibrio fischeri in Tom Knight’s lab.

It surprisingly easy to grow bioluminescent microbes. I’ve found a couple of undergraduate-level protocols online for isolating marine Vibrio species from fresh squid. I’m currently in the middle of my second attempt at isolation. These are my notes so far. I’ve adapted this Basic Protocol for Isolating Bioluminescent microbes for use in my kitchen.

I’m using food-grade agar I got from a food-hacking friend, powdered chocolate jello (I ran out of powdered agar), Baking Soda instead of CaCO3 (chalk), and canned tuna water instead of powdered LB broth (hope it doesn’t have preservatives in it).

13 Jan 2010

18:30 Got fresh, “uncleaned” squid from New Deal Fish Market in Union Square: $2.50

18:55: got 1L gatorade

Added 1/2 teaspoon of sea salt to the empty gatorade bottle

22:00 cut squid head in 1/2, added to tap water in gatorade bottle

14 Jan 2010

22:00 It GLOWS! wow!

took pictures with ad-hoc cardboard camera frame; cutting gatorade bottle in half (squid now in what resembles a petri dish with 8-inch tall walls)

Added 15 mL of Jeff’s Food-Grade agar to 450 mL of water; not enough? Still liquid. (only have a 15 mL falcon tube, not a scale).

Boiling the agar liquid in a water bath. Will add Jiffy CornMeal to supplement.

22:25 agar solution at 175 C. Cools to a mildly viscous liquid.

22:30 adding 15 mL chocolate JELLO

How many grams is 15 mL of agar powder? 0.34 g / cm^3 (http://www.wolframalpha.com/input/?i=15+mL+Agar) 15 mL ~ 5g.

How about Jello? 1.1 g / cm^3 (http://www.wolframalpha.com/input/?i=15+mL+Jello) 15 mL ~ 17g

Most LB plate protocols call for 15 grams of agar powder / Liter. I started with 5g (15 mL) in 450 mL of water. So I was 2/3 short.

It turns out 15 mL Jello + 15 mL agar in 450 mL boiling water solidifies into something approximating an agar plate. I’m adding 2.5 mL more Jello to thicken it a bit.

Other ingredients of recommended for Luminescent Agar plates (per 450 mL H2O):

• 2.5 g CaCO3 (NaHCO3 is 2.173 g/cm3; and I’m going to substitute it. Added ~ 1g)
• 5 g Glycerol (left it at sprout)
• 15 g NaCl (2.165 g/cm3, added 4 mL; ~ 8.66 g)
• 4 g “dehydrated Nutrient Broth” (I’m using a teaspoon of liquid from a can of tuna)

0:00 15 Jan 2010 Poured “Luminescent Chocolate Agar”

0:30 poked bright blue bioluminescent patches on the squid carcass with a toothpick and streaked them out on the choco-agar.

Later that summer, I read Steven Levy’s “Hackers: The Heros of the Computer Revolution.”

Now I’ve started to see an analogy between the tools used by nascent biotech hackers today and the computer hackers of the late ’70s.

The upshot of improvements in diagnostic kit technology are more than just clinical: cheaper, faster, broader ways of interrogating the natural world will be a boon to everyone interested in understanding it. In particular, I believe the amateur / non-institutional biotechnology community requires easy, low-cost methods for asking questions and getting answers about biological systems, and my intuition tells me a lot of those methods will be based on diagnostic kit technology. To me, the impact of future diagnostic kit technology on amateur biotechnologists will be roughly analogous to the impact microcomputer kits had on the “amateur computer scientists” of the late 1970’s and early ’80s.

One of the first things we looked at was the trending terms on twitter; Specifically, for every day over the last 6 months. I made some fun aggregate visualizations of this with whiskerplots using topfunky’s sparklines gem in ruby: here’s the top 100 trends, and here’s the top 2000. (here’s the code.)

8 twitter trends from 1-jan-2009 to 5-jun-2009

That was all weeks ago, though. Today is a happy day for the WEP because today, we published our first report: Iranian Election and Twitter: The First Eighteen Days. I only contributed a little bit to some of the early twitter stats (I’ll write more about them later) – David Fisher and others were the real hard workers.

From June 7 – 26, we recorded 2,024,166 tweets about the election in Iran, and we found out some pretty interesting things. Guess how many of those 2 million tweets were retweets! Go check out the report to find out (pdf).

Iranian Election and Twitter: The First Eighteen Days

sampling sites from The Marine Viromes of Four Oceanic Regions - we can do a better job of mapping than this!

Then I’d like to do something a little more abstract. The first thing that comes to mind is distorting the map so that one of the dimensions of the data, such as read density (base pairs/ km^sq), is constant for each pixel. This would expand the map in areas that had been sampled and shrink it in areas that were not sampled. I actually think read density is a pretty boring thing to graph (at least until there are tens of thousands of samples), but I think this technique would be a neat way to represent something like a diversity metric for each sample. These kind of visualizations are called “cartograms.”

(Note: these are just some basic ideas. It might be the case that all density representations, such as heatmaps or the distortions I mentioned above, are inappropriate for representing sparse samples across a large area. Nonetheless, the first step is to get some data and start experimenting.)

In this cartogram “the sizes of states are proportional to the frequency of their appearance in news stories.” From Diffusion-based method for producing density-equalizing maps by Michael T. Gastner and M. E. J. Newman.

a tab-delimited example of a basic metagenomic dataset for constructing a map:

sample_id   lat lon sequence_id 16s_sequence    suspected_species
000001  32.131341   98.231332   0001    agcctagcacgga...    Bacillus subtillis
000001  32.131341   98.231332   0002    agcgtaggttgac...    Acinetobacter baylyi

I would be happy just with 10,000 entries in a single text file in a format similar to the one above (but note that lat/lon are identical for all sequences in a given sample). It would be even better if there were more dimensions of data. Here are some other potential columns:

• taxonomy (calculate a diversity metric from each sample from this? what else can we do with a taxonomy?)
• pathogenicity
• auto- or heterotrophic
• other metabolic information?
• GO terms, or something like them at an organismal level
• URL canonical species description in ncbi
• ? Please make suggestions in the comments.

Synthesizing such a dataset (as a large plaintext file or as a database) will require aggregating a variety of other datasets. I have no idea where to begin with them. If I know a particular species (Acinetobacer baylyi, for instance), is there a single entry point for deriving all this information in NCBI?

existing metagenomics datasets

I spent a couple of hours reading metagenomic papers and browsing around for datasets. Here’s a quick list of interesting resources. My naive first look didn’t turn up anything similar to the basic plaintext example above.

MEGAN – Metagenome Analysis Software & sample data

Methods for comparative metagenomics (introducing MEGAN) (paper)

UniFrac software & sample data (Look for the datasets they used to construct the phylogeny trees)

UniFrac paper (paper)

The Marine Viromes of Four Oceanic Regions (paper)

Data from Marine Viromes study (and more!)

ncbi metagenomics book soil chapter (Waseca County Farm Soil)

16s rDNA identified from Waseca County Farm Soil dataset (in ncbi’s nt database)

CAMERA: A Community Resource for Metagenomics (their webview of the datafiles looks interesting; the datafiles themselves are just fasta)

CAMERA: Surface Water Marine Microbial Community Gene Expression project

inspirational infographics:

Travel-time Maps by Chris Lightfoot & Tom Steinberg

Center for Mathematical Modeling infographics by Juan Pablo De Gregorio

• Temperature Gradient Gel Electrophoresis (TGGE) and Denaturing Gradient Gel Electrophoresis (DGGE)

DGGE of small ribosomal subunit coding genes was first described by Gerard Muyzer, while he was Post-doc at Leiden University, and has become a widely used technique in microbial ecology. PCR amplification of DNA extracted from mixed microbial communities with PCR primers specific for 16S rRNA gene fragments of Bacteria and Archaea, and 18S rRNA gene fragments of Eukaryotes results in mixtures of PCR products. Because these amplicons all have the same length, they cannot be separated from each other by agarose gel electrophoresis. However, sequence variations (i.e. differences in GC content and distribution) between different microbial rRNAs result in different denaturation properties of these DNA molecules. Hence, DGGE banding patterns can be used to visualize variations in microbial genetic diversity and provide a rough estimate of the richness and abundance of predominant microbial community members.

Solid-phase PCR (Polony PCR) is also intriguing: Dilute and spread the DNAs across a surface and use solid-phase PCR to amplify individual DNAs into polonies. Pick polonies and have them sequenced.

• In situ localized amplification and contact replication of many individual DNA molecules

“We describe a method to clone and amplify DNA by performing the polymerase chain reaction (PCR) in a thin polyacrylamide film poured on a glass microscope slide. The polyacrylamide matrix retards the diffusion of the linear DNA molecules so that the amplification products remain localized near their respective templates. At the end of the reaction, a number of PCR colonies, or ‘polonies’, have formed, each one grown from a single template molecule. As many as 5 million clones can be amplified in parallel on a single slide. If an Acrydite modification is included at the 5[prime] end of one of the primers, the amplified DNA will be covalently attached to the polyacrylamide matrix, allowing further enzymatic manipulations to be performed on all clones simul-taneously. We describe techniques to make replicas of these polony slides, and high throughput sequencing protocols for this technology.”

• More info: Solid phase DNA amplification: characterisation of primer attachment and amplification mechanisms
  

Lastly, I could just give up on the idea of a heterologous PCR product and instead design primers specific to a particular rDNA or other gene of interest. Instead of exploring the space of all rDNAs present in the sample’s metagenome, I would be testing for the presence of a particular DNA. Much less exciting, if you ask me.

I’m reading about Unit Tests and Zope and Test Driven Development. I am going to read the wikipedia page on each of the concepts and probably some other pages elsewhere on the web. I want to be able to highlight the one or two most valuable snippets from each page and use those together as a the history entry.

When I go to my history list, I don’t really care about the list of pages I visited, I care about what I was looking for and why I went there.

I want task- and concept-centric history lists.

notes: Generalize the concept of a content consumption stream, a content history. What if it magically existed for all the books and papers I read in college. Would each page be an entry? Each paragraph? Each book or article? That wouldn’t be as useful as some kind of… distillation of the key concepts:

• acting as pointer to help me find my way back to content surrounding main article (a bookmark)
• and as a stand-alone distillation or note of the record

In the pilot program, hundreds of participants will sign up at bioweathermap.org, gather in a “FlashLab” event, receive a sampling kit (probably something like a Q-tip and a test tube), and disperse throughout the local region to sample a particular type object, such as cross-walk buttons. The participants will then return their samples to the BioWeatherMap group for analysis.

Once several hundred samples have been collected and prepared, the BioWeatherMap group will purchase a single DNA sequencing run on a high-throughput DNA sequencer for all the samples at once, with each sample receiving about 1500 “reads”. The preparation step isolates and amplifies a small, species-specific region of DNA from the genome of each sample. For each sample, then, up to 1500 unique species could be identified (or the same specie could be identified 1500 times). By leveraging economies of scale, the BioWeatherMap group will be able to provide 1500 reads at the most economical cost.

The BioWeatherMap project is the first example of a new kind of distributed, participatory, bite-size science. It will demonstrate to participants that research doesn’t require a Ph.D. The data they generate will be fascinating – literal BioWeatherMaps: maps of time-series of microbial population flows overlaid on a Google Earth view of the city.

The first Flashlab event should be happening sometime this summer (2009).

In the meantime, I want to figure out how to do a smaller-scale version of the same thing – direct sequencing of a short, species-specific genomic DNA sequence for microbial species identification – in a cheap, garage-friendly way. I want to do Garage Metagenomics. (DNA Barcoding is the same concept but for Eukaryotic organisms).

I’ve started looking through the metagenomics literature(links to fulltext library) for simple protocols that could be adapted to a basic garage lab. I’m planning on outsourcing the actual sequencing ($50-$100?), but doing the rest of the sample preparation myself: isolating and purifying genomic DNA and doing PCR to amplify the species-specific DNA barcode, probably a 16s or 18s ribosomal subunit gene.

Check back soon for more info on my progress and please leave a comment if you would like to help out or do some Garage Metagenomics of your own.

Unfortunately, I dislike reading PDFs on my computer screen – I would like to access the library on a smaller, more portable, preferably e-ink device. I have a kindle 1 and an OLPC, both of which have the smaller form factor and tablet mode that I like. Unfortunately, the kindle 1 does not support PDF and my attempts at converting journal articles and textbooks have been unsatisfactory and unreliable at best.

OLPC vs. Kindle 1

I would love to use the kindle, since it can provide ~8000 pages on a single 3-hour battery, is very light, and has a great screen, but I would be limited to whatever I could fit on an SD card (SDHC is unreliable but may work), and if I had converted the PDFs into images, I would lose the ability to search the content.

The OLPC seems more promising. It’s not light enough to comfortably hold for long periods of time, but I usually can find a way to prop it up on a knee or table. The battery only lasts 4 hours or so on a full charge, but it can natively display PDFs in tablet mode and has a remarkable screen that approaches e-ink in clarity (although not viewing angle). It can run debian & gnome, and I feel like there has got to be a good application out there for that environment built for managing large document collections and searching across them. It has like 2GB of built-in flash memory for the OS and an SDHC slot.

I’ve taken some photos comparing the screens of the OLPC natively showing a PDF and of the Kindle 1 showing a .prc file of the same PDF converted with PDFRead.

Basically, I want iTunes for PDFs on Linux. With fast full-text search.

Oh, I have my eye on the Kindle DX and the FoxIt eSlick e-ink readers. Both can display PDFs natively.

Here’s a short screencast I made demonstrating how I like interact with my pdf library.

(Download the source of this video from vimeo for more detail – it’s only 60 megs and is 1280×720)

Part 1: Spatiotemporal Transcriptional Logic

Consider the marvelous spatiotemporal program an organism’s genome must encode to organize embryogenesis and development.  Cells divide and specialize, divide and specialize, recursively executing the developmental program stored in the genome according variables mediated by time and space.

This process is reliable and repeatable: for instance, (under normal growth conditions), wild-type C. elegens develops into precisely 959 somatic cells, for each of which the lineage and developmental fate is known.   The implications continue to amaze me – is development really so deterministic?!

Davidson College’s Genomics class introduced the notion of spatiotemporal logic with an in-depth analysis of the genetic control regions of a Sea Urchin gene called Endo16.  Eric Davidson’s lab essentially spent years reverse engineering the genetic logic controlling Endo16’s expression in developing Sea Urchin embryos.  They elucidated the identity of modular patterns of regulatory DNA upstream of Endo16.  These regulatory modules act as inputs for a variety of transcription factors.  To understand the transcriptional logic encoded by the pattern of regulatory modules, the Davidson team quantitatively measured the effect different transcription factor combinations had on the output of the regulatory region i.e. expression of Endo16.   The logic they deduced was so coherent they were able to write it in pseudocode.

Here is a diagram from one of their papers describing the transcriptional logic:

Let me quote from the introduction of their paper:

Each gene, in each cell of a developing animal, must read and respond to the presence or absence of multiple inputs. In effect, these inputs provide the gene with the regulatory information it requires to determine its own activity: this includes signaling inputs from adjacent cells and inputs that indicate what other relevant genes have been functioning in the cell in which the gene resides. These inputs are presented to the gene in terms of concentrations and activities of nuclear transcription factors. The heritable structural basis for cis-regulatory information processing functions consists of the target site sequences at which transcription factors bind to the DNA. The identity and disposition of these sites specify the regulatory activities that can be executed by the cis-regulatory system, depending on circumstances. This genetic hardwiring causally determines to which inputs each gene regulatory system will (Davidson, 1990; Davidson, 1999; Davidson, 2001).

So essentially, the spatial and temporal variation in transcription factors are the inputs for logic gates controlling gene expression.  Sound interesting?  The next step is to check out The regulatory genome and the computer, a review the Davidson team wrote in 2007 that is in ways the genomics version of Von Neumann’s The Computer and the Brain.  I’ll leave you with the abstract:

The definitive feature of the many thousand cis-regulatory control modules in an animal genome is their information processing capability. These modules are “wired” together in large networks that control major processes such as development; they constitute “genomic computers.” Each control module receives multiple inputs in the form of the incident transcription factors which bind to them. The functions they execute upon these inputs can be reduced to basic AND, OR and NOT logic functions, which are also the unit logic functions of electronic computers. Here we consider the operating principles of the genomic computer, the product of evolution, in comparison to those of electronic computers. For example, in the genomic computer intra-machine communication occurs by means of diffusion (of transcription factors), while in electronic computers it occurs by electron transit along pre-organized wires. There follow fundamental differences in design principle in respect to the meaning of time, speed, multiplicity of processors, memory, robustness of computation and hardware and software. The genomic computer controls spatial gene expression in the development of the body plan, and its appearance in remote evolutionary time must be considered to have been a founding requirement for animal grade life.