Choosing Biostats PhD: Michigan vs UPenn?

[Primadonna]

Anyone have any insight or advice for choosing between these PhD programs for someone whose research interest is statistical genetics/genomics? Michigan's reputation with genetics and Penn's association with Perelman School of Medicine is what draws me to both programs. Penn professors say they work directly with MDs and doctors in the School of Medicine and Michigan is just a School of Public Health in comparison, but Michigan historically has the better reputation and ranking and overall seems to have a lot of connections in the field. Michigan has more faculty working in genetics but I'd be part of a much bigger cohort and probably have to fight harder to work with a specific faculty. I also got accepted for a training grant at Michigan. Anyone want to drop any advice please? Location/area/environment/etc. isn't a huge determinant for me.

[bayessays]

If you're on the training grant, I doubt you'll have to worry about finding faculty to work with, and besides, Michigan has so many statistical genetics people that I'm sure there will be multiple you'd be interested in anyways.

What's the advantage in your mind of working in the medical school instead of the public health school?  I don't think this should be a big factor unless you're looking to marry a doctor and want to be near them.

If you are serious about doing statistical genetics, I don't think there is much of a decision to be made here - go to Michigan.

[Primadonna]

13 minutes ago, bayessays said:

If you're on the training grant, I doubt you'll have to worry about finding faculty to work with, and besides, Michigan has so many statistical genetics people that I'm sure there will be multiple you'd be interested in anyways.

What's the advantage in your mind of working in the medical school instead of the public health school?  I don't think this should be a big factor unless you're looking to marry a doctor and want to be near them.

If you are serious about doing statistical genetics, I don't think there is much of a decision to be made here - go to Michigan.

Every time I met with someone from Penn they pushed the med school as a big deal. They work directly with a seemingly never-ending supply of doctors and MDs at Perelman and they see their work have a stronger impact "than you could ever see at a school of public health" (roughly). Guess it's not as big a deal as they make it out to be? ?

[bayessays]

That's crazy.  Who cares if you work with MDs if you want to be a statistician or geneticist?  The people at Michigan are regularly publishing in Nature and you'll start working with them day 1.  You're going to spend your first two years at Penn doing laboratory rotations.

[bayessays]

If you wanted to work on clinical trials or something like that, they may have a point, and that type of research experience and the rotations at Penn may be a strength for a certain subset of people, but not you.

[trynagetby]

4 hours ago, bayessays said:

I don't think this should be a big factor unless you're looking to marry a doctor and want to be near them.

Marrying a Upenn Med School doctor while in PhD could effectively quintuple your stipend, so I would seriously consider this perk if you got game.

In all seriousness though, if you're sure you want to do genetics Michigan is the best program hands down. I've talked to friends at JHU Biostats, visited Umich Biostat as part of an REU, and talked to Profs at Harvard biostats visit day. Every one of them mentioned Michigan Bio-statistics Department as being the dominant program in statistical genetics.

About working with Doctors in Med-School and such, I wouldn't worry about it as the only benefit of this is having lots of data and potential problems to solve. It's clear professors in Michigan profs don't worry about finding data/problems/collaborations and Michigan itself has a very good medical school. If you like super applied work, than maybe Upenn might be better?

Michigan's center for statistical genetics is at  the point where people nationally come to them to solve the big problems in genetics rather than just "how do I analyze this study". For example Michael Boehnke pretty much mapped out the major genetic pathway for type II diabetes (i'm too biologically challenged to understand what this means and not sure if i'm saying it right, but the thing he did related to this was apparently a big breakthrough in genetics in the 21st century).

Edited April 7, 2021 by trynagetby

[cyberwulf]

University of Michigan also (checks notes) has a medical school, and I'm sure plenty of Biostat faculty collaborate with them. And while UM Med (#15 tie on USNews) isn't Penn (#9), it's still very, very good, so there isn't going to be a discernible gap in the quality and availability of medical collaborators. Penn Biostat being under the med school may have a few modest advantages, but it also has some drawbacks. Biostat is a massive academic outlier in the context of the medical school, so there won't be nearly the same kind of cohort experience that you'd get by interacting with students with overlapping research interests studying epidemiology, health services, etc. in a school of public health. Also, access to public health researchers expands the set of collaborative projects that Biostat students may get engaged with, and often these projects provide great(er) fodder for methodological development.

[StatsG0d]

17 hours ago, Primadonna said:

their work have a stronger impact "than you could ever see at a school of public health" (roughly)

Two thoughts:

1. I'm a graduate student in biostat at a SPH and I have personally collaborated with several doctors.
    
2. That statement sounds sort of arrogant to me. It's preposterous to think that the work done in medical schools is more impactful than that in public health schools, particularly considering the pandemic. IMO, it's a pretty exciting time to be in a public health school.

[sugarbeet]

Michigan’s stat genetics program is historically strong, but their faculty’s research is a little out of date in recent years. Most of the people there are still working on genome-wide association studies. For more general biostatistics, SPH is fine, but for genetics/genomics research, being in the school of medicine will make a difference as the problems that they work on are more significant. 

[StatsG0d]

14 hours ago, sugarbeet said:

For more general biostatistics, SPH is fine, but for genetics/genomics research, being in the school of medicine will make a difference as the problems that they work on are more significant. 

IMO the fundamental difference would be Med School researchers are typically focused more on applications, whereas SPH is more focused on methods. 

"More significant"... let's check

• Michigan Faculty
  • Goncalo Abecasis 230,000+ citations
  • Michael Boehnke 140,000+ citations
  • Hyun Min Kang 46,000+ citations
  • (stopping here--there are many more)

The only prof I can find at Perelman on Google that does work in Statistical Genetics is Hongzhe Li (25,000+ citations). I'm sure there are more, but since is the most mentioned AND is the PI for the Lab of Statistical Genetics and Genomics, it's safe to say that he's one of if not the best researchers there, and yet has far fewer citations than the truncated list mentioned above.

A note on rigor:

Michigan requires a course BIOS801--Advanced Inference, which discusses asymptotic theory and other advanced topics. As far as I can tell, there are no such advanced courses required at UPenn.

[bayessays]

If anything, the Michigan people are doing work more on the applied end.  Hongzhe Li seems to be doing work on the more statistical end, publishing in journals like JASA and Biometrika.  If you want to work on genetics problems but be more of a traditional statistician as well, that is something to take into consideration.  There is not a lot of that from the genetics people at Michigan - they're publishing in Nature, Human Genetics, etc.  If you do statistical genetics at Michigan, you will basically become an actual trained geneticist as well as statistician, not a statistician who happens to work on genetics problems.

[StatsG0d]

Boehnke has a couple Statistics in Medicine papers in the last few years. Min Kang has a few Bioinformatics papers. Overall, it is true that statistical genetics is a fundamentally applied area of research. But even new methods in genetics don't frequently get published in traditionally stats / biostats journals, typically because they have a relatively specific audience (namely, geneticists and doctors).

[sugarbeet]

4 hours ago, StatsG0d said:

IMO the fundamental difference would be Med School researchers are typically focused more on applications, whereas SPH is more focused on methods. 

"More significant"... let's check

• Michigan Faculty
  • Goncalo Abecasis 230,000+ citations
  • Michael Boehnke 140,000+ citations
  • Hyun Min Kang 46,000+ citations
  • (stopping here--there are many more)

The only prof I can find at Perelman on Google that does work in Statistical Genetics is Hongzhe Li (25,000+ citations). I'm sure there are more, but since is the most mentioned AND is the PI for the Lab of Statistical Genetics and Genomics, it's safe to say that he's one of if not the best researchers there, and yet has far fewer citations than the truncated list mentioned above.

A note on rigor:

Michigan requires a course BIOS801--Advanced Inference, which discusses asymptotic theory and other advanced topics. As far as I can tell, there are no such advanced courses required at UPenn.

Goncalo Abecasis spends most of his time in industry now. Those high citation papers are large consortium applied papers that aren't necessarily good for PhD students who need to get methods papers for graduation. I have heard some students struggled to graduate. 

[sugarbeet]

31 minutes ago, StatsG0d said:

Boehnke has a couple Statistics in Medicine papers in the last few years. Min Kang has a few Bioinformatics papers. Overall, it is true that statistical genetics is a fundamentally applied area of research. But even new methods in genetics don't frequently get published in traditionally stats / biostats journals, typically because they have a relatively specific audience (namely, geneticists and doctors).

It is true. The audience of genetics/genomics methods is different from general biostatistics. Top genetics/genomics methods papers are typically published in journals like Nature Methods, Nature Communications, Genome Biology, and Genome Research etc.

[StatsG0d]

15 hours ago, sugarbeet said:

It is true. The audience of genetics/genomics methods is different from general biostatistics. Top genetics/genomics methods papers are typically published in journals like Nature Methods, Nature Communications, Genome Biology, and Genome Research etc.

Right.

There may be some differences, but I feel like the vast majority of biostat grad students and faculty consider Michigan to be the best place in the country (if not the world) for statistical genetics by a relatively large margin. I simply wanted to point out that it's folly to say that their work isn't as impactful as what's done in a med school. At least measured by citations, which I feel is a decent proxy for impact, it's hard to beat the faculty at Michigan. 

[bayessays]

@sugarbeetmakes some good points about Michigan.  Their genetics grads do take longer to get their degrees (6 years is pretty much the minimum, I think).

[sugarbeet]

3 hours ago, StatsG0d said:

Right.

There may be some differences, but I feel like the vast majority of biostat grad students and faculty consider Michigan to be the best place in the country (if not the world) for statistical genetics by a relatively large margin. I simply wanted to point out that it's folly to say that their work isn't as impactful as what's done in a med school. At least measured by citations, which I feel is a decent proxy for impact, it's hard to beat the faculty at Michigan. 

Michigan definitely has the best reputation for stat genetics, which is largely based upon Mike Boehnke and Goncalo Abecasis's reputation. Their trainees have got faculty positions at places like Harvard, UPenn, and Emory etc. That is mostly when genome-wide association studies were hot. But now that the field has already moved on to newer technologies, traditional stat genetics is no longer that attractive. In fact, their recent trainees can hardly get faculty positions. Now that Goncalo Abecasis is going to industry and Mike Boehnke is close to the age of retirement, I am not sure that Michigan's other stat genetics faculty are necessarily stronger than other places. 

[sugarbeet]

2 hours ago, bayessays said:

@sugarbeetmakes some good points about Michigan.  Their genetics grads do take longer to get their degrees (6 years is pretty much the minimum, I think).

>6 years is crazy! most Biostatistics students finish in less than 5 years.

[bayessays]

Just now, sugarbeet said:

>6 years is crazy! most Biostatistics students finish in less than 5 years.

The genetics training grant requires you to take extra genetics coursework that takes some time. 

All great points to bring up though. I still think the variety of younger faculty like Zhou, Jiang, Wen, and some of their newer hires I'm not familiar with make it an appealing option for someone who wants a lot of options in the statgen world.  But of course there are also other factors to consider in making a decision.  If @Primadonna is even a little uncertain of whether they want to do statistical genetics, opting for the training grant and Michigan would be a much less appealing option.

[ciabatta]

I applied to UPenn, but was rejected ? I really liked their lab rotations. I was uncertain about what to study so their lab rotations is a huge draw to me, but unfortunately I didn't get in. 

[Primadonna]

3 hours ago, ciabatta said:

I applied to UPenn, but was rejected ? I really liked their lab rotations. I was uncertain about what to study so their lab rotations is a huge draw to me, but unfortunately I didn't get in. 

I was ready to accept my Michigan offer today, but then I come back and check this thread and I have doubts again. I really liked the lab rotation idea too at Penn, although there's more faculty at Michigan I'm interested in than Penn.

 

4 hours ago, bayessays said:

The genetics training grant requires you to take extra genetics coursework that takes some time. 

All great points to bring up though. I still think the variety of younger faculty like Zhou, Jiang, Wen, and some of their newer hires I'm not familiar with make it an appealing option for someone who wants a lot of options in the statgen world.  But of course there are also other factors to consider in making a decision.  If @Primadonna is even a little uncertain of whether they want to do statistical genetics, opting for the training grant and Michigan would be a much less appealing option.

I don't really have any uncertainty about wanting to work in genetics, I'd be really, really surprised if I winded up doing something else. I'm a little concerned about Michigan's program taking significantly longer than Penn's. Most Penn students finish in 4-5 years from a Bachelor's. I knew that Michigan's program would take longer but I didn't know that genetics students at Michigan took 6+ years. I've also heard qualms before this thread that Michigan's genetics reputation is on the decline. The first few posts convinced me that wasn't true but now I'm unsure again.

[ciabatta]

2 minutes ago, Primadonna said:

I was ready to accept my Michigan offer today, but then I come back and check this thread and I have doubts again. I really liked the lab rotation idea too at Penn, although there's more faculty at Michigan I'm interested in than Penn.

 

I don't really have any uncertainty about wanting to work in genetics, I'd be really, really surprised if I winded up doing something else. I'm a little concerned about Michigan's program taking significantly longer than Penn's. Most Penn students finish in 4-5 years from a Bachelor's. I knew that Michigan's program would take longer but I didn't know that genetics students at Michigan took 6+ years. I've also heard qualms before this thread that Michigan's genetics reputation is on the decline. The first few posts convinced me that wasn't true but now I'm unsure again.

I have no experience with genetics, but I know people who does genetics at UPenn Biostats and seems pretty happy. They seem to work on cool problems in single cells and use machine learning a lot. I was so disappointed that I didn't get in. 

[Primadonna]

4 minutes ago, ciabatta said:

I have no experience with genetics, but I know people who does genetics at UPenn Biostats and seems pretty happy. They seem to work on cool problems in single cells and use machine learning a lot. I was so disappointed that I didn't get in. 

I'm sorry to hear that. ☹️ Did you apply anywhere else? Where did you wind up selecting to attend?

[ciabatta]

2 minutes ago, Primadonna said:

I'm sorry to hear that. ☹️ Did you apply anywhere else? Where did you wind up selecting to attend?

I will go to Rutgers. I wanted to stay on the East coast.

[Primadonna]

1 minute ago, ciabatta said:

I will go to Rutgers. I wanted to stay on the East coast.

That's still a great choice. Location doesn't matter too much to me, and I think I'd prefer a smaller city like Ann Arbor than Philadelphia, but that's not a deal-breaker for me.