• My first year I was an overwhelmed young grad student with a poster and the silly idea that I could pack my schedule full of sessions all day every day without suffering melting of the brain. (You really need to pick and choose at PAG. It’s like an all you can eat buffet of science, it is all to easy to go overboard.)
• My second year I returned to PAG as an actual presenter giving two talks to packed sessions (which isn’t an endorsement of my own science I was sandwiched between successful scientists who also happened to be gifted speakers both times).
• And now in my third year I’ll get to see PAG through the eyes of an exhibitor. No, this doesn’t represent my post-PhD career path. This year PAG happened to fall in the break between filling my dissertation and the start of my next “real” job.

Anyway, my plane is about to board so I should wrap this up. To all the rest of you who are coming to the conference, hope you have a great conference, don’t push yourselves too hard, and drop me a line if you’d like me to hook you up with a free t-shirt.

My transformation obviously isn’t complete yet though. Lab meetings with pizza sounds like a wonderful idea.

So it gives me great pleasure to report that, as of December 14th (last Friday), I have reached that goal.

Behold! The lollipop handed to every newly minted Berkeley PhD when their thesis is accepted.

What was my thesis about you ask? Well I still don’t have a good elevator speech, so let me simply say that the first part of my thesis has to do with how plant genomes change over time and the second part demonstrated a new method for learning the function of pieces of DNA which don’t code for proteins but instead determine where and when neighboring genes will be turned on or off.

So what’s next? This whole site traces its origins back to travel posts I put up to let friends and family know how I was doing as I interviewed as various graduate schools. So I suppose there would be a fair bit of symmetry to shutting it down as I leave grad school, but I don’t want to do that. Now that I’m finished with my PhD, I’m looking forward to rediscovering the things I used to do for fun, and I remember writing updates here used to be a lot of fun.

On a more practical level, what comes next for me is a 2000 mile drive from California to the midwest (with all my worldly possessions packed into the back of my car) to visit family for the holidays. I am suddenly very conscious of the fact I haven’t driven on snow in more than four years. After that it’ll be onward to a post-doc.

If you’ve left an unanswered comment in the last six months or so and are still interested in me getting back to you, let me know.

For now… it is good to be back.

There are two completely different steps to reconstructing maize subgenomes: 1) putting together ancestral chromosome pairs 2) grouping one copy of each ancestral chromosome together into subgenome 1 and the other copy of each ancestral subgenome 2.

Ancestral chromosome pair reconstruction:

This step hinges on a single assertion backed up by a number of papers: inversions and other forms of within chromosome rearrangements significantly outnumber translocations (movement of sequence from one chromosome to another). This observation has been backed up by a number of previous studies.

Give that fact (within chromosome rearrangements are more common than between chromosome rearrangments), when two different pieces of the same maize chromosome are orthologous to different parts of the same sorghum chromosome, we can say that those parts of the maize chromosome originated from the same copy of that sorghum chromosome in the twenty chromosome (two equivalent to each single chromosome of sorghum) tetraploid ancestor of maize.

A good example of this is the comparison between maize chromosome 1 and sorghum chromosome 1 in Figure 1 of the paper you are inquiring about. There are actually two separate segments of maize chromosome 1 which are orthologous to different parts of sorghum chromosome 1. In theory each segment could have come from a different copy sorghum chromosome 1 in the tetraploid maize ancestor and were combined on one maize chromosome through translacations, but the most parsimonious explanation is that both segments came from a single maize chromosome.

Once we reassemble that first copy of sorghum chromosome 1 (the two pieces on maize chromosome 1), we know that any remaining pieces of the maize genome orthologous to sorghum chromosome 1 must have come from the second copy of that chromosome in the tetraploid ancestor of maize. In this case there is one piece on maize chromosome 5 and one piece on maize chromosome 9, but since we have already reconstructed one whole copy of the chromosome, the only source for these two segments is the second chromosome copy.

Repeat this logic 9 more times and you have reconstructed ten pairs of ancestral maize chromosomes, just like we did in the paper. (And as cited in our PNAS paper, we weren’t the first to come up with this idea. Before the maize genome was even complete or the sorghum genome was published, another group was able to derive approximately the same ten chromosome pairs by comparing a genetic map of maize to the rice genome: http://www.plosgenetics.org/article/info%3Adoi%2F10.1371%2Fjournal.pgen.0030123 )

Assigning ancestral chromosomes to subgenomes

In this step we examined the number of maize-sorghum orthologs shared between each maize ancestral chromosome and the orthologous sorghum chromosome. This is figure 2 from our 2011 PNAS paper. As you can see in that figure, for each pair of maize ancestral chromosomes, one copy retains a lower percentage of genes found in sorghum along its entire length and the other copy retains a higher percentage of genes found in sorghum along its entire length.  (Too avoid being thrown off by new/transposed genes in sorghum we excluded sorghum genes without orthologs in rice, but you can get the same results without that filtering step, the main difference is that the % retained is lower.)

The maize ancestral chromosome copy which retained a higher percentage of genes also found in sorghum was assigned to maize subgenome 1 and the maize ancestral chromosome copy which retained a smaller percentage of genes also found in sorghum  was assigned to maize subgenome 2.

Then we went back and colored maize subgenome 1 blue in figures 1 & 2 and maize subgenome 2 red in those same two figures. Which I think is one of the things that confuses people. The red and blue color coding in Figure 1 shows a distinction we didn’t actually make until after the analysis shown in Figure 2.

Schnable, James C., Nathan M. Springer, and Michael Freeling. “Differentiation of the Maize Subgenomes by Genome Dominance and Both Ancient and Ongoing Gene Loss.” Proceedings of the National Academy of Sciences 108, no. 10 (March 8, 2011): 4069 –4074.

It comes from setting unrealistically fast deadlines for yourself. And then meeting them.

Sometimes that means working early mornings, late nights, and weekends. Sometimes it means coming up with a new approach, getting the results in three hours, and sneaking out of lab at 3:30. But the point is the results are what matter. If you can find ways to be unexpectedly productive you’re much less likely to burn out entirely than if you can only ever meet your own deadlines by burning the midnight oil at both ends (mixed metaphor intended).

Working hard for the sake of appearing to work hard (either to others or to yourself) is the surest road to burnout and lack of results.

P.S. Productivity goes up at least 5-fold when not also teaching.

P.P.S. If the reagents you are working with are as old as you are, you need to worry. (That falls into the working hard but not getting results category.)

And all I can say is….

what a rush! This is why I love what I do for a living. Two days of improvising and lit-searching and throwing different approaches against the wall to see what would stick. And at in the last 24 hours I finally managed to turn my proposal into a project I would actually enjoy carrying out.

The only problem is that now I kind of want to spend next weekend doing the same thing. Ideally with a shot at actually getting some cash if I successfully sold people on the value of my research. It’s been a couple of months but I’ve finally been re-bitten by the science bug! Speaking of which, I should wrap this up. My alarm is set for 7 AM tomorrow so I can get to lab in time to squeeze in an RNA extraction before class. I’m taking yet another shot at building a proper sequencing library. Wish me luck!

But when you work, you do awesome things.

Sex specific splicing of a gene of unknown function of a gene syntenically conserved in all grass species.

With only four days work I was able to go from a giant pile of reads (from the still not properly appreciated Davidson 2011 The Plant Genome) to figures like the one above.

So what is the figure above showing you? One of a large number of genes which show a different pattern of splicing in male and female reproductive organs in maize.* The region “E8″ is usually treated as exonic in female reproductive tissues but is spliced out like an intron in male reproductive tissues. What does it mean (if anything)? I have no idea yet! But it would have been a real pain to try to re-invent the wheel for identifying these deferentially spliced genes in python. In R, once I figured out the right incantation, it’s practically plug and play for any gene you could possibly be interested in. Including the software for the (actually quite useful) visualization shown above.

So thank you R. What you do — once I can figure out how to make you do it — you do incredibly well.

*Maize makes it easy for us by separating female and male flowers into two entirely different organs (the ear and tassel respectively).

Data from:

Analyzed using the R package DEXSeq:

Anders S, Reyes A, Huber W. 2012 Detecting differential usage of exons from RNA-Seq Data. Unpublished. (Link is to a PDF)

• Downloading, decompressing and quality/adapter trimming more than 800 million RNA-seq reads (four full Hiseq 2000 lanes).
• Attempting to make my very own transcriptome assembly for a species where the genome is available but doesn’t look to be published anytime soon.
• Figuring out how to look at differential use of exons in maize between male and female floral structures.  (Later on this will involve using some R packages. I’m not looking forward to that part. R always makes me feel like I’m coding with one hand tied behind my back).

The surprising part is that I’m not being held up by a lack of processors to throw at the problem (the usual problem in computation work), nor a limited supply of RAM (probably the biggest problem in bioinformatics specifically). Instead I’m hitting the limit of how fast all these various programs can read data off of hard drives and write results back. Right now I am waiting for a little surplus capacity to free up.

It’s hard to believe that eight months from now this will all be over.  I started my education back in 1990. If they kept numbering years in school after high school I’d be a 20th grader right now. But my adviser has informed me that I need to have graduated by this December, so that’s what I have to make happen. Next week is my last as a graduate student instructor. This summer and part of the fall will be a mad sprint to finish up various projects and collaborations and get them written up for publications, then thesis writing, signing, and submitting are all that stand between me and (hopefully) the last degree I’ll ever need to earn.

As you can see golden plant2 is expressed at high levels in photosynthetic tissues and not expressed at all in tissues like roots, endosperm, and pollen. Do you know how long it would have taken me to profile the plant-wide expression pattern of a gene this comprehensively by isolating RNA from different tissues using qPCR? WEEKS! Do you know how long it took for me to get the same level of insight with qTeller? 90 seconds!

Do you know how long it’ll take you to regenerate this same analysis? 30 seconds. Just click this link. There have been so many awesome RNA-seq papers coming out recently for maize. I know when I arrive in Portland on Thursday for the Maize Genetics Conference I’m going to see a whole lot more even bigger/better RNA-seq datasets which people haven’t finished writing up yet. Some of these datasets have been on posters since the very first maize meeting I went to back in 2009 when I was a wide-eyed first year and may _never_ get published.* But others will be published weeks or months from now, making this visualization all the more powerful.

But for now, MORE EYE CANDY:

Anther ear1

Expression of anther ear1, a mutant in the gibberellic acid biosynthetic pathway

Link to regenerate analysis

Glossy1

Link to regenerate analysis. 

Glossy1 mutants change the type of wax produced on the leaves of developing maize seedlings, so it makes sense that the gene shows high expression in both maize seedlings and mature leaves. I can even sort of explain away the high expression in developing seeds and embryos since the the primordia which will eventually become the first leaves of the next generation of corn plants are beginning to form, But why in the world does glossy1 show such high levels of expression in anthers?

*Here is the relevant excerpt from my previous rant on data analysis:

I recently did the math on a PLoS Genetics paper published in late 2009 based on on a single in-depth analysis of RNA-seq comparisons of mutant and non-mutant siblings. Today we could generate the same dataset, with twice the depth of sequencing for less than $1000 dollars. (INCLUDING regent costs). The takeaway lesson here: just because your dataset was expensive to generate doesn’t mean you don’t have to worry about the competition stealing the glory if you take more than a year to publish. 

A) Suzzy the grad student is mapping a recessive mutant which makes the pollen of cornplants shrivel up and die. By examining a bunch of known genetic markers in plants with dead pollen and normal pollen producing siblings of those plants she has narrowed the location of the gene responsible for her trait down to a region of only a couple of megabases on the fifth chromosome of maize. Since the whole maize genome contains over 2,300 megabases of sequence that means she’s already ruled out 99.9% of the genome. But her region still contains, say, a dozen genes and she needs to know which one she should check first to see if mutation in it is responsible for her mutant phenotype.

B) Johnny is another grad student. He wants to understand how corn plants genetically regulate how wide their leaves will grow to be. By measuring a lot of plants descended from two parents, each with known genotypes, he can identify regions of the genome where inheriting information from one parent or the other seems to be correlated with either wider or narrower leaves. He calls these regions quantitative trait loci (or QTLs). Now he has picked the genetic region that seems to have the biggest effect, and he wants to know what gene within the region is actually responsible for the effect.

There are a number of ways for both Johnny and Suzzy to narrow down their lists to the genes most likely responsible for the changes they are each observing in corn plants:

1.  A okay candidate might contain a protein domain with a function related to the plant trait they were studying.
2. A good candidate might show a specific pattern expression in the part of the plant where they observe their phenotype (pollen for Suzzy and developing leaves for Johnny).
3. A GREAT candidate would be a gene which has already been studied by another group and has a mutant phenotype related to the phenotype currently observed.*

Checking out the protein domains of genes under a QTL or mutant interval can be accomplished in almost any genome browser (for example the ones provided by MaizeGDB, or CoGe).

Checking the expression patterns of genes is harder. One option for plants is a website called PlexDB which lets you look up the expression of individual genes in different people’s microarray experiments. The biggest problem with microarrays is that it really is impossible to compare expression between different people’s microarray experiments.

Identifying known mutant genes within a genomic interval used to be a real pain. Now they could search the classical maize gene list to see if any of the genes in their interval appear on it, but it is still not what you’d call efficient.

Now imagine a website that let you check all those things in one step!

Example of data on genes in a region 10 megabases from the start of maize chromosome1. Click to zoom in and make the text big enough to read.

The automated annotations of protein domains are the same ones you’d find in any of the genome browsers above, but instead of checking each gene in turn, qTeller reports them all in a handy sortable spreadsheet (which you can either view online or download to your computer).

It also incorporates measurements of gene expression using RNA-seq data from papers published by the maize community.** To make the numbers as comparable as possible I went back to the raw reads provided by NCBI’s Sequence Read Archive and taken each dataset through the same analytical pipeline.

And, of course, it reports any classical maize genes which lie within the interval a researcher is studying (taken from the version two classical maize gene list.)

Plus syntenic orthologs in other species. Because, at least in the grasses, genes with mutant phenotypes are disproportionately likely to have been retained at the same location in the genome of lots of different grass species.

Suzie found a gene within her interval which was expressed at much higher levels in pollen than in leaf tissue:

And Johnny realized that his QTL contained the mutant gene milkweed pod1, a classical maize mutant known from previously published papers to be involved regulating in leaf development.

So yeah, qTeller is what I’ve been working on for the past few months. Please let me know if you run into any bugs or have any questions.  -James

*Admittedly this would be rather disappointing news for the grad students involved (it’s a lot harder to get a splashy paper out of rediscovering a known mutant), but it’s much better to find out you might be studying a known gene early so you can check and cut your losses if it turns out to be true, instead of after you sink another couple of years into recloning and recharacterizing it.

**The authors of all these papers deserve a whole bunch of credit for generating these datasets:

• Waters AJ, Makarevitch I, Eichten SR, Swanson-Wagner RA, Yeh C-T, et al. (2011) Parent-of-Origin Effects on Gene Expression and DNA Methylation in the Maize Endosperm. The Plant Cell doi:10.1105/tpc.111.092668.
• Davidson RM, Hansey CN, Gowda M, Childs KL, Lin H, et. al. (2011) Utility of RNA Sequencing for Analysis of Maize Reproductive Transcriptomes. The Plant Genome4:191-203 doi:10.3835/plantgenome2011.05.0015
• Li, P., Ponnala, L., Gandotra, N., Wang, L., Si, Y., et al. (2010) The developmental dynamics of the maize leaf transcriptome. Nature Genetics 42: 1060-1067. doi:10.1038/ng.703
• Wang, X., Elling, A.A., Li, X., Li, N., Peng, Z., et al. (2009) Genome-Wide and Organ-Specific Landscapes of Epigenetic Modifications and Their Relationships to mRNA and Small RNA Transcriptomes in Maize. Plant Cell 21: 1053-1069. doi:10.1105/tpc.109.065714
• Jia, Y., Lisch, D.R., Ohtsu, K., Scanlon, M.J., Nettleton, D., et al. (2009) Loss of RNA Dependent RNA Polymerase 2 (RDR2) Function Causes Widespread and Unexpected Changes in the Expression of Transposons, Genes, and 24-nt Small RNAs. PLoS Genet 5: e1000737. doi:10.1371/journal.pgen.1000737
• The Maize Gametophyte Project: Unpublished Dataset SRP006965