Lab Core values. 

  1. Each lab member has a unique perspective and skill set, and it is the job of the PI to adapt their leadership and mentoring approach to put the lab member in the best position to succeed. We prioritize long term trajectories over short term productivity.
  2. We let knowledge or capability gaps drive our research, rather than searching for applications of our research post hoc. If we do not have the required expertise to address an important gap, we invest in what is needed to obtain it. If the gap is addressed, we build upon this advance and address the next gap, onward ad infinitum.
  3. We prioritize principled approaches that allow us to infer models of underlying biology, rather than heuristic approaches evaluated according to predictive or classification accuracy. 
  4. When the goals of a project are clear, we develop infrastructure that can be repurposed and built upon, rather than ad hoc solutions. When the goals are unclear, we develop proof-of-concept solutions that can be rapidly iterated until the goals become clear.

Why you should do a postdoc (or, what to learn at each stage of your career).

The degree of training a researcher should pursue depends on their long-term career goals. One general piece of advice is that, once you have developed or been trained with a certain skill set, you can always choose not to use it or to take a job that doesn’t require it. But, the converse is not true.
PI Flannick often describes the various stages of training as follows:

  1. As an undergraduate, you learn how to learn. You learn how to (with minimal oversight) listen to lectures, take notes, read textbooks, work through problems, and ask questions.
  2. As a masters student, you learn more about a specific topic. This is essentially a continuation of the undergraduate experience, albeit at a more advanced and specialized level.
  3. As a PhD student, you learn how to conduct a research project – that is, to solve a problem with no known answer. You learn how to transform vague ideas (provided by your PI) into something concrete that can be published.
  4. As a postdoc, you learn how to define a research program. You understand a field of study, learn what vague ideas are worth pursuing, and learn how to string vague ideas into progress over a 5 year period.
  5. As a faculty member, you learn how to enable other people to execute your research program. You learn how to raise money, establish a lab culture, and ensure people are set up for short and long term success.

If your goal is to be a faculty member, most people really do benefit from a postdoc – even if you could get a faculty job right out of a PhD, there is a degree of seasoning that you learn by giving yourself time to establish a more complete research program. If your goal is to be a research scientist or to go into industry, the postdoc is probably not necessary – most likely there will be a PI (or manager) you work for that will provide you with things to work on. In this setting, even a PhD may not be beneficial. If you would like to be given vague problems to work on (and a fair amount of freedom to solve them), you should get a PhD. If you are okay (throughout your career) being given specific tasks to work on, you don't need a PhD.

Mentoring philosophy.

Our main mentoring philosophy is that there is no one philosophy. Every trainee and every researcher has different preferences in terms of how much guidance and oversight they want, how comfortable they are with asking for help, and the specificity of instructions they require. Therefore, we usually favor an adaptive approach where the first assignment for any trainee is carefully chosen but described at a relatively high-level. The expectation is that the trainee should struggle with this, formulate questions, and then come to the PI for help. Ensuring that the PI has time for regular meetings is critical for this to work, but it is also critical that the trainee is open about what they do and do not understand, rather than trying to “impress” the PI with how little help they need (you would be surprised at how often this failure mode occurs!). It is usually pretty clear after a few meetings what mentoring style works for the trainee. Some trainees do great with weekly meetings and minimal guidance. It is more common though for trainees to benefit from at least two meetings a week plus conversations over Slack in the interim – the most common failure mode for junior trainees is working an incredible amount but running in the wrong direction. See the previous topic above regarding what you learn as a PhD/postdoc!

Picking a lab.

By far and away, the most important feature of your lab is your fit with the PI – meaning, their philosophy regarding training and how to approach science. The second most important feature is your fit with other lab members – do they complement your skills and provide you with an opportunity to learn more things? After all, you will spend most of your time working with them.

The skills you will learn in the lab are probably the third most important aspect of it – do you want to become an expert in a disease area? A specific methodology? Or learn a skill set (i.e. statistical modeling or data science)?

The specific area of research is probably number four. PI Flannick did his PhD in microbial network alignment (after knowing no biology before beginning!) and switched during his postdoc to statistical genetics. It is not nearly as hard to transition to new areas of research as you might think. What is hard is to establish the proper foundation for how to do science and, secondarily, to become an expert in a specific technique of methodology. Those two skills will be your competitive advantage, whatever area of research you focus on during the rest of your career.

Bioinformatics, computational biology, statistical genetics, and genomic analysis.

Our lab conducts work in all of these areas, which are sometimes used interchangeably.

  1. Bioinformatic projects focus on building (usually cloud-based) infrastructure to efficiently store data, run computations across them, and serve results publicly. Most of the work maintaining our knowledge portals falls into this bin. The primary skill set needed/developed by these projects is software engineering.
  2. Computational biology projects focus on computing novel things from biomedical (in our case, usually genetic and genomic) datasets. Computational biology incorporates more algorithmic development than software engineering, and therefore develops/uses skills from computer science (which is distinct from software engineering).
  3. Statistical genetic projects focus on modeling and inferring biological parameters from genetic and genomic datasets. They differ from computational biology projects primarily in their greater dependence on statistics and statistical inference. These projects develop both Bayesian and frequentist statistical knowledge.
  4. Genomic analysis focus on analyzing datasets to answer biological questions, by applying bioinformatic, computational biology, and statistical genetic techniques. To succeed in this area, it is not necessary to be able to develop new software, algorithms, or statistical methods, but it is necessary to be familiar with cutting-edge methods and the questions they address, have good organizational skills, and be able to “get things done”. This area develops/uses skills from data science. The skills to be successful in this area are often not fully appreciated (particularly by method developers) – it is one thing to analyze data to evaluate a method; it is a whole other thing entirely to learn how to tell a biologically compelling story given all of the data and tools at your disposal.

What about machine learning?

Despite the prominence AI/ML plays in tech and popular culture, thus far our lab is somewhat bearish on its applicability to genomics. There are some clear clinical opportunities for it, for example in patient diagnosis or monitoring. But, in terms of scientific applications, the issue is that (a) we are usually lacking the large-scale training data required to fit these models and (b) we usually care about inferring things about the state of nature, rather than simply building a good classifier or predictor. Genomics is a “big data” science, but the “big data” is usually well upstream of what we analyze – by the time petabytes of read-level sequence data reach us, they have usually been distilled through numerous analytical pipelines to produce summary statistics and representations that are orders of magnitude smaller. These representations are actually getting pretty big themselves, but we usually analyze them to infer parameters about disease or genetic architecture rather than to predict or classify, for which statistical inference is often a better technique. In spirit, statistical inference is very similar to machine learning, and at any rate it is probably better for you in the long term to learn how to construct and fit statistical models rather than simply apply machine learning techniques.

Having said all of that, the new generative LLM models do have potentially interesting applications to genomics, and we have some early-stage projects in that space.

Picking a project in the lab.

The first step is to determine whether you want to develop skills in software engineering/bioinformatics, computational biology/algorithm development, statistical methodology and inference, or genetic analysis/data science. The second step is to be realistic about your current skill set (although new lab members usually worry too much about this). We have had successful trainees who have entered the lab with no computational experience – for them, beginning with an analysis/data science project enables them to start working with data and gradually move “up the chain” toward method development and/or implementation if it suits them (see the topic above regarding the four types of projects in the lab). It is more common for new lab members to have no biological experience; they typically begin with software engineering or computational biology projects and then move “down the chain” toward data analysis.

We usually have multiple trainees and staff working on each type of project, who are there to either help train lab members who want to gain experience in a particular area, or to collaborate and provide methods/analyses to lab members who do not.

Whatever your interests, experience, and trajectory, it is the job of the PI to pick a project that appropriately balances “quick wins” and the possibility for early  publication with “high upside” and possibility to develop a long-term research program.

BCH, HMS, the Broad Institute, and Boston.

Our lab has affiliations with three organizations, which can be confusing for those not familiar with the Boston biomedical/biotech landscape.

BCH is the institute that employs the PI, and which employs many/most of the trainees. Our academic/research grants are primarily at BCH. Our division is Genetics and Genomics, which focuses primarily on rare disease research. BCH is one of the hospitals affiliated with HMS. 

HMS is our academic institution, through which we recruit students. Most faculty members at BCH (or any of the other HMS-affiliated hospitals) have academic appointments at HMS. HMS doesn’t provide any lab space for us, however, and we don’t apply for grants through HMS – those are through BCH and the Broad.

The Broad Institute is an independent non-profit organization with which we are affiliated. Most faculty at HMS, MIT, or Harvard (which is distinct from HMS) can obtain an Associate Member appointment at the Broad. This allows us to (a) attend all Broad community events; (b) apply for grants at the Broad; and (c) have office space at the Broad. Most of our portal grants are through the Broad, and most of our research staff and software engineers are employed by the Broad.

Trainees in the lab therefore receive three appointments: at BCH, at HMS, and at Broad. These appointments provide IT and building access at each location. Our lab has office space at both BCH and Broad.

BCH is based at Longwood, as is HMS and most of the HMS teaching hospitals. The Broad is based in Kendall Square. Longwood has more of a medical and academic vibe, whereas Kendall has more of a genomic and biotech vibe. Both are valuable to be exposed to.

Is the HMS/Broad the right place for me?

To get the downsides out of the way first, the HMS/Broad research community can be a little overwhelming, and it is very unlike a traditional academic institution at a university. There is a very high faculty to trainee ratio, a much higher preponderance of research staff, and a lot more funding. It is a “big research” type of place.

Having said that, if you are motivated to learn about and conduct impactful research, there is really no place in the world like the HMS/Broad community. There are word-class experts in biology, human disease, genomics, and computer science everywhere. As a trainee, you will meet and collaborate with a surprisingly large number of them. You will be exposed to an unparalleled diversity of perspectives, areas of expertise, and career stages. It is one of the best places in the world to become a true scientist and to learn how to conduct scientific research.

Is your lab the right place for me? 

Our lab has a unique balance of access to big projects and big data, funded through our knowledge portal grants, and access to the PI (who is still at a relatively early career stage). The answers to the questions above should give you a sense as to what we value, what we work on, and how we think about the world of scientific research. 

How to begin. 

Contact us! The first step would be a conversation in which we learn a little more about you and your goals and jointly determine whether the lab would be a good fit for you. We are always seeking talented and motivated new members!