Data Carpentry hackathon for genomics

I'm pleased to report back from the Data Carpentry (DC) genomics hackathon, which I attended last week with ~26 other folks at Cold Spring Harbor Labs in New York. The goal of this meeting was to develop modules for a DC workshop focused on analysis of next-generation sequencing and other genomic data. The original DC lessons were designed for a very general audience using ecological data, so we were tasked with outlining, organizing, and starting to write materials for a two-day workshop specifically for genomics.

Each of the following points could be thoroughly explored in their own post, but here are a few highlights from this meeting:

• Attendees were a great mix of biology researchers and educators from a range of institutions (research intensive, primarily undergraduate), computer scientists, and assessment specialists. This meant we were pulling from a broad range of skills, and incorporating multiple perspectives in planning.

• The length of the meeting (2.5 days) allowed us to get a running start on actually developing materials (GitHub repos here prefaced with "genomics"). In addition to "intro to Unix" material that would largely remain constant from the original DC lesson, we started developing six modules that cover a general genomics workflow: setting up a project, getting to know your data, data wrangling (QC and alignment), analysis and visualization, and cloud/HPC. I personally found it remarkable and gratifying to see so much attention paid to the initial preparatory stages of a project.

• Numerous folks emphasized the importance of understanding your target audience. Some of these discussions related to the assumed skill level (or pre-requisites) for workshop attendees. Other conversations related the need to accommodate particular cultural or gender issues while teaching to make the learning environment comfortable for everyone. 

• What makes DC workshops special and distinct from other courses? In developing the modules described above, we talked about the distinction between Software Carpentry and Data Carpentry, as well as if and when instructors should be expected to teach about biology (rather than computing/data analysis). The general consensus is that the focus of DC on telling a narrative about data means we should be emphasizing "best practices" for improving productivity and reproducibility, rather than advocating for particular types of analyses. That being said, there is ample opportunity during lessons to model rigorous methods, as well as provide extra resources for students to improve their skills in experimental design and statistical reasoning.

• A particularly challenging aspect of developing such resources is assessment of student improvement following a workshop. It's challenging to evaluate how much students will retain after such a short period of time (2 days), as well as whether these skills will transfer over to their research methods. One breakout group focused on developing a strategy for surveying students prior to and directly following a workshop to measure immediate learning, as well as 3-6 months following to measure long-term gains. We targeted question formats that would address student learning in terms of the following areas: declarative knowledge (Can you recall this fact?), skills (Can you write this code?), and attitude (Will you use this skill?).

I was initially on the fence about whether to apply for the hackathon. I'm a first year professor wallowing in the murky depths of teaching a new course, and my overtaxed brain was whispering that maybe it would cause too much stress. My gut, thankfully, doesn't always listen to my brain. Moreover, the class I'm piloting this semester is an undergraduate bioinformatics class focused on genomics, so the DC hackathon fit naturally into my preparation for the last few weeks of the semester. I'm looking forward to reporting back soon about my semester-long class is wrapping up, as well as my first teaching experience for Software Carpentry workshop in a few weeks.

Developing a bioinformatics class as a part of a new bioinformatics/genomics minor.

Since I've been waxing philosophical about the small tasks I've been performing for students this semester, it seemed opportune to talk a bit more in depth about my big job for the semester: designing an undergraduate bioinformatics lecture and lab for next semester (spring 2015). This is actually something I'm really excited about, so I'm willing to expend quite a bit of time and energy being mindful and deliberate in how I set up the class schedule and assessments.

While I figure out particulars about hosting my own website, I've started a research webpage through Google Sites as a stand-in. This page includes resources related to a new undergraduate minor in Bioinformatics and Genomics created by my department chair, Srini Kambhampati. I'll eventually end up teaching genomics as well, but am mostly focused on developing the bioinformatics class material right now. I've got lots to say about the general premise of bioinformatics/computational biology education, but will save that for a later date.

I've talked to lots of folks about how they teach (or would like to teach) bioinformatics. Some are deeply rooted in a particular programming language (like perl), or focused on a particular type of bioinformatics (i.e., next generation sequencing analysis). I'm trying for a wider approach to capture a variety of student interests. Here's my first shot at explaining what my class will be like.

Considerations:

1. Students will represent a mix of levels of computational and biological expertise, from folks who are rather computer-phobic but biologically savvy to computer science folks with relatively little biological knowledge. Lots of other biological research here at UTT is ecologically focused, and there's a lot of health professionals on campus and in the community
2. I want to leverage extant and familiar educational technology (Blackboard, GUIs) while introducing students to more efficient and useful interfaces (Github, CLI, etc).
3. I'm not expecting students to be expert programmers in any particular language at the end of the semester (with the exception of unix shell scripting, which is a necessary baseline for functionality).

Primary course objectives:

Students should be able to:

1. Describe the scope of bioinformatics research and applications (including but not limited to next-generation sequencing, comparative genomics, ecological informatics, metagenomics, extensions to health informatics, etc)
2. Design and implement bioinformatics pipelines to answer pre-defined questions from a variety of biological disciplines.
3. Validate results from bioinformatics algorithms using hypothesis testing, correcting for multiple comparisons, etc.
4. Characterize the limitations of data to answer questions of interests.
5. Obtain resources to learn new languages and algorithms (i.e., be self-directed in learning more on their own).

How I'm doing it (which is mostly a list of questions I need to answer):

1. In addition to the quick overview pages on my research website (which I'm hoping to transition over to the UTT Biology page when they are completed), I've set up a Github repository for the class that includes a modular syllabus (following the guidelines here) and will eventually include the materials used for in-class (lab) exercises. I'll still be posting grades on Blackboard, and haven't decided how to share lecture materials and assessments (Github, Blackboard, Slideshare, Wiki?).
2. I'm not requiring students purchase a textbook. I will be relying on published scientific literature for objective 1, and (free to download!) OpenIntro Statistics for objectives 2-4 (in addition to examples from published literature). Of course, objective 5 teaches them to find and evaluate resources on their own.
3. UTT Biology has a new building addition which will be completed by the end of this semester, and I've been put in charge of getting the computer lab set up. I'm hoping class sizes will be small enough that each student can work on their own machine. I'm still trying to decide if and how to include HPC/cluster usage, given that most analyses can run locally if the appropriate software is installed. There is the opportunity to get an instructional allocation from TACC, and I know other folks who have used Amazon Web Services Cloud Computing.
4. I'm visiting classes taught by my colleagues this semester (both in computer science and biology) which may include students who are interested in taking my class next semester. I'm hoping to get a better idea of student desires and educational background to be able to craft materials that are both relevant and useful to them. I'm also considering cross-listing the class at the graduate level, as a few folks have asked me about that possibility (Sidenote: I'm brainstorming ideas for graduate level classes on my research page, too).

If you have ideas or experiences that you'd care to share, I'd be happy to hear them! Stayed tuned here, on my research webpage, and the Github repo over the next few months to see how the class develops.

Bladderwort revisited: Because the fun never ends in Junk DNA Land!

I've had some feedback and read additional commentary posted elsewhere about the bladderwort news I posted about earlier this week and feel compelled to post a follow-up. I also just submitted revisions to a manuscript, which means I'm validated in indulging my whims of...continued professional development via blogging?

Anyway, I particularly like Jonathan Eisen's Twisted Tree of Life award to MSNBC's coverage of the bladderwort study. His argument is compelling and his rhetoric is golden.

Next, my friend Jared astutely pointed out in a comment on my last post that I appear to have misrepresented the research from the ENCODE project. Again, here I fell victim to a common folly among transposable element researchers. I've adopted the term "junk DNA" to refer to any non-coding sequence in a genome, whereas others use the term in a functional sense (i.e., junk DNA is a sequence to which function has not been ascribed). I've since edited the previous post to reflect my viewpoint. Simply put, a large portion of the human genome is comprised of transposable elements and other non-protein coding sequence, yet contemporary research indicates many of these sequences still perform functions necessary to organism function. I appreciate the comment, Jared, and am so grateful for folks like you who help me stay on my toes!

Finally, there was also my embarrassing gaffe in mistyping "Badderwort" in my title of the original post. It's since been corrected, but will live on in the automatically generated permanent web address. It's ok, though, because the bladderwort is definitely the baddest of the worts regardless.

Bladderwort: tiny plant, tiny genome, weird name. So what?

My dad emailed me a link this morning. Then a NESCent coworker passed along another. Then my boyfriend tagged me in a post linking a similar article on Facebook.

All those news articles refer to a recent paper (Ibarra-Laclette et al., 2013, Nature) which reports on a newly sequenced plant genome (Utricularia gibba). YAWN, we all say, it's a silly little plant and we can't even eat! What makes it news worthy?

Turns out there are a few reasons why. First, it's called badderwort, and that's just funny. Second, bladderworts are a wickedly awesome type of carnivorous plant which capture prey by rapidly sucking them into bladder-like traps (complete with a trapdoor!).

Then there are a whole slew of other (more "publishable-in-Nature") type reasons to care about bladderworts. This species has a remarkably small genome but still possesses the same number of genes as other plants surveyed. Most of the news articles linked above talk about the implications of the research for our understanding of "junk DNA," or portions of the genome which do not code for protein sequences. It's admittedly a hot topic in science news right now, following heavy on the heels of research in human genomics suggesting that a large portion of the human genome is junk (EDIT: but still functional). The repercussions of the bladderwort genome research is that junk DNA isn't really necessary for "healthy organisms," as we now have an example of one species which has managed to "throw out" junk DNA.

I find the bladderwort research interesting for a few reasons. First, my professional research interests revolve around junk DNA and how changes in genome size affect organisms. The main focus of my research here at NESCent is comparative genomics of a group of plants called Asparagales (includes agave, onion, iris, orchid, asparagus). Species in Asparagales have genomes which range from very small (i.e., a few times larger than bladderwort) to ginormous (200,000 times larger); the variation in genome size is like due at least in part to junk DNA. The conclusions from the bladderwort news articles are kind of no-brainers: OF COURSE plants can persist quite happily without tons of junk DNA. There are fungal species which entirely lack junk DNA, too.

I'm also pretty tired of the semantics surrounding the "Is junk DNA important?" argument, which has been very well described elsewhere (and there's concomitant throwback from the creationist community as well, but that's for another day). To me, junk DNA is just another source of mutation. In the context of evolutionary biology, junk DNA provides more opportunity for evolutionary innovation. That's a hypothesis which is better tested with a broad sampling of species' genomes to infer changes over time. A snapshot of a single genome at a single point in time can't tell you as much as broad comparisons. Asking about the relevance of junk DNA to an organism living today is like asking whether money is required for humans to survive. In a strict sense, currency ISN'T required for humans to persist. It IS a means to an end, though, and a mechanism through which contemporary humans purchase food, shelter, and other resources required to survive. That might be a crappy analogy.

A better question to ask from the bladderwort research is what pathways allow for regulation of junk DNA and other mechanisms of genome downsizing. It's not just about the ability of the bladderwort genome to remove sequences already present. It's also about the ability of the genome to prevent other sequences from incorporating themselves into extant sequences, since some junk DNA is derived from viruses and the genome can be continually invaded. There is also evidence the bladderwort genome has undergone whole genome duplications at least three times, but has managed to remove most of the extra genetic material (including extra genes!) after each event. Now THAT'S some cool science.

Architecture and evolution of a minute plant genome, Ibarra-Laclette et al., Nature (2013) doi:10.1038/nature12132

Catalyzing landscape genomics.

Earlier this month, I participated in a catalysis meeting at NESCent on landscape genomics. There is a common theme for each catalysis meeting. Sometimes that uniting principle is a historic area of research with multitudes of literature that could benefit from synthesis and identification of uniting principles. Sometimes the theme is an emerging area of science for which new questions are being elucidated. The landscape genomics catalysis meeting was interesting in that it addressed several major themes in evolutionary biology (population genetics, signatures of selection, comparative genomics) and was looking for ways to unite them theoretically and statistically into predictive methods. I was initially curious why the organizers chose to invite me, but quickly came realize how aligned the themes of the meeting are with my own research interests. Here are a few impressions from the time I spent sharing my thoughts with other scientists.

First, catalysis meetings are an interesting conglomeration of scientists. The goal of such meetings are to throw 25-30 scientists from different research foci and at various stages in their careers into a room together.  No one person was an expert on landscape genomics. There were experts in landscape ecology, comparative genomics and statistics, and we worked together to devise common projects and goals. The beautiful part of synthesis is that contributions from various participants are required to achieve the goals of the meeting. There are many people contributing, and everyone learns something. As a recent article from NESCentians points out, this is a fertile ground for incubating new collaborations, projects, and ideas.

Second, an aggregate of such brains allows new insight into how science works. Well, that's certainly a vague statement. At this meeting in particular, several of us were struck by the differences in how genomics and ecology view data. The former tends to throw out data left and right, paring down sequences and levels of variation so the remaining data can be described more easily. Ecologists, on the other hand, attempt to describe the overall variation in a given system, and try to model the nature of that variation. I personally believe genomicists have a great deal yet to learn from ecologists in this respect.

All in all, I believe this meeting was particularly fruitful. Sometimes tempers and egos flare at these meetings, to the point that I'm urged to start chanting "Fight, fight, fight!" and hope for fisticuffs. As exciting as it is to see other scientists so passionate about their research, I was pleased that members of this meeting were more focused on bridging gaps between disparate disciplines. There are a number of incipient projects jumpstarted from it, including one manuscript I volunteered to lead.

Synthesis FTW!