Mar 202012

After twice reading The Emperor of All Maladies: A Biography of Cancer last spring, I was determined to focus on cancer research in the future. Since then, I have been struggling to find the right cancer type to work on. Cancer is a deep field, and it takes a lot of efforts to become a real expert on any one cancer type. The ideal cancer for us to work on should satisfy the following criteria: 1) has high incidence and death rate in both US and China; 2) DF/HCC has demonstrated research excellence such as a SPORE program; 3) I am able to find good clinical collaborators in China.

Most of my previous collaborations especially with Myles Brown and Nelly Polyak have given us a footing in breast and prostate cancer research, although these two cancers (especially prostate cancer) are not as bad in China. DFCI is famous for leukemia and lymphoma research because of Sydney Farber and we have started collaborating with Jon Aster and Stephen Blacklow. We could also potentially collaborate with the Ruijin Hospital in Shanghai which has the best leukemia research in China and are also world renowned for their acute promyelocytic leukemia (APL) treatment. GI cancers, including gastric, liver, and colon cancers, are very serious in China. Epigenetic profiling and drug deliver for liver cancer is the easiest among all tumors, but DFCI is not known for liver cancer research. Lung cancer is probably the biggest cancer killer in both US and China. DFCI has an excellent lung cancer program and we could also collaborate with the Shanghai Pulmonary Hospital, although I previously had zero background in lung cancer research.

A couple of weeks ago, someone from Bill Kaelin’s laboratory at DFCI approached us for a collaboration. I started reading Kaelin’s website and papers. He has been amazingly productive with just a medium-sized lab. I know from my seminar visits that many of his former trainees are successful at various faculty positions, so he must be a very nice person and a supportive mentor. His pioneering work, such as that on E2F1, Rb, VHL, HIF, and EglN, has made quite an impact in cancer research. Interestingly, he is not focused on specific cancer types, but has been focused on tumor suppressors that function in multiple cancers, especially those that are related to cancer metabolism. This is a fairly new area for me, but both cancer metabolism and tumor suppressors are very exciting and promising. Maybe if we focus on cancer epigenetics, we don’t have to pick a specific cancer now, and could learn the different cancers on the way. Of course, it would take a lot of hard work and probably Kaelin’s genius (as told by his postdoc). At least I look forward to working with this group and learning a lot about cancer biology from them.

Mar 162012

When I started graduate school at Stanford, I talked to all the faculty in the BMI program about research and rotation. I planned to do four rotations, one in each quarter. One faculty I was interested in trying was Geo Wiederhold, but he only considered rotation students with at least a six-month commitment working with him. I didn’t want to cut my rotation from 4 to 2, so decided against a rotation with him. Over the years, as bioinformatics becomes a more mature discipline, I started to appreciate Geo’s wisdom. Bioinformatics is an interdisciplinary field at the intersection of biology, statistics, and computer science. It actually takes about one year for most of the students to learn enough to really do anything useful.

Harvard is a tough place for PhD students. It is not unusual for a laboratory to have 10 postdocs but only 0-1 graduate student. Most of the postdocs in the lab are too busy to teach beginning PhD students (especially not for a year), so only the very well prepared, truly smart and motivated students could thrive here. At the Jan DFCI strategic planning, Matt Meyerson said something profound, “I don’t train new people in my lab, but the infrastructure in the lab train the new people”. This would suggest that laboratories that are fairly big (~30 people) with all levels of trainees (graduate students and postdocs) and staff (technicians and research scientists) would actually be better for graduate students. Their size requires people of different levels, so will have a good infrastructure to train new graduate students.

Now at Tongji we have a fairly big laboratory with people of different levels (assistant professors, postdocs, different years of PhD, MS, and undergraduate students). It will be impossible if all the new students coming to the lab only learn from the PIs, therefore we need to create the infrastructure to train them. Here is a draft plan:

  1. Two-week summer internship, on RNA-seq and ChIP-seq. Mostly test students’ general capabilities and EQ.
  2. Six-week fall training, finish 6 HW / projects for Harvard STAT215 course. More systematic training, with professor lectures, wider topics, more HW on dealing with real data and serious programming.
  3. 6-12 month infrastructure training based on student level and progress. Students will be asked to process high throughput datasets or write programs. This stage will get students to be familiar with running bioinformatics algorithms, visualizing and interpreting results, developing better sense of data, and build stronger programming skills. These not only help our existing projects, but also build students’ general bioinformatics skillset, and some might result in publishable work.
  4. Towards the end of the infrastructure training, students could help out on research projects of other lab members or start their own research project.

We need designated students and faculty to setup this infrastructure, so we can train better students, postdocs, and junior faculty.

Mar 082012
  1. Common interest and complementary expertise between the experimental and computational groups.
  2. Combine collaboration and methodology for the computational biologists.
  3. Co-authored papers, some last authored by the experimental PI and others last authored by the computational PI.
  4. One designated experimental student/postdoc paired with one designated computational student/postdoc on the project.
  5. Regular meetings between the experimental and computational students/postdocs (e.g. weekly or even more frequently), and also with the experimental and computational PIs (e.g. monthly).
  6. Experimental and computational groups have established expertise on the respective domain of question this collaboration project is on.
  7. Computational biologists are able to learn a lot and get useful feedback from the experimental collaborators.
  8. Experimental PIs and postdocs are knowledgable and passionate about the collaboration project.
  9. Good project design, have some idea of a potential hypothesis and followup plans early on.
  10. Joint grants to ensure long-term commitment.
Mar 072012

Many computational biologists use a Mac (desktop or laptop) for research. In addition to the standard software suites such as Microsoft Office and Adobe, which your department should have a site license, I have found the following software to be very good in improving my productivity.

  1. Skype: Members of my lab stay on Skype when they work, and it is very convenient to brainstorm and discuss ideas over Skype, especially when I travel.

  2. DropBox: We use this to share files, which is extremely useful when a group of people are revising a paper together.

  3. Evernote: This is wonderful for taking notes. It allows me to generate multiple notebooks and organize them under different categories.

  4. GrandProspective: Very quickly scan folders to see their disk usage. Helps me find useless files / folders that are taking big chunk of my HD.

  5. Apimac Timer: We use it to make sure our lab meetings stay in time.

  6. Coconut battery: Keeps track of my Mac’s age and its battery capacity.

  7. IGV: Local genome browser to look at NGS data, really powerful!

  8. *1Password*: Keep tracks of the many online passwords in one place. It is easy to transfer the password file when I move to a new computer.

  9. *Papers2*: Great for keeping on track with recent publications in the community. Nice viewing and note taking functions. I can also drag manuscript pdf to Papers, which will parse out the reference information, and it can format references for manuscripts.

  10. *Things*: I use Mac Mail, which doesn’t have a very good to-do function as in Outlook. Things is a good to-do list management. I use calendar for events I know the absolute time/date, but Things for events only with a deadline but not a specified “get it done” time.

* are not free, but they are worth the license fee.