This post was meant to be just a brief snapshot aimed at students wondering where bioinformatics is going in 2008 and beyond. What’s the future of bioinformatics? What kind of focus should you develop in the near future? What kinds of skills will you need?

More after the jump, below.

He writes:
If one were to ask the people of virtually any African village (outside some 10 countries devastated by AIDS) what their greatest concerns are, the answer would undoubtedly be the less sensational but more ubiquitous ravages of hunger, dirty water and environmental devastation. The real-world needs of Africans struggling to survive should not continue to be subsumed by the favorite causes du jour of well-meaning yet often uninformed Western donors.

Google is creating an archive of science and tech lectures online. There are some gems on Google video and related sites.

Related sites are also linked to, below.

Some of my favorites:

Hans Bethe’s physics lectures makes quantum theory accessible to anyone on Quantum Physics made Relatively Simple:
I suggest everyone take some time to read about his life on the “About” tab if you are not familiar with the man.

Interviews and lectures by Richard Feynman=

There are also some good lectures on biological sequence analysis, genomes, and semantic web (for instance) through Google.
Some bioinformatics-related videos

Some more off-the-beaten path physics:

A lecture on supersymmetry made at Burning Man 2006

Many of us decided we’d still be scientists or mathematicians, but do research that would never rely on a grant application again.

At least one of us decided we’d open up a commune or institute and invite in all the students and scientists who had interesting ideas, but who would never get a chance to get funding, much like the FQXi is doing these days (and what a good idea).

Still others said they’d leave science forever and just wander around the world.

I asked the table, why couldn’t we do both? Wander the world, do what we want, and still do science, if we had a lot of money? Several people almost immediately said that it was unusual for anyone to expect to ever publish any ideas in any journals without an affiliation or a set-up lab of some kind. The big-brother-aspect of affiliation was an important thing to consider. The impression was that where you publish from may be closely related to where you publish.

Which brings me to the buzz that’s surrounding e8 and Garrett Lisi’s work on a unified theory of everything. My last educational excursion into theoretical physics was over a decade ago, so I’m not qualified to comment on the details, problems, and successes of his unified theory across e8. But, he and I have some things in common: we have apparently gone to some of the same Burning Man festivals, we both share an interest in physics, and we’re generally non-traditional types of people. I don’t feel uncomfortable commenting on that.

I believe that a valuable contribution that Dr. Lisi is offering the world is the view that someone can live outside the mainstream and still love science and “do science”. Dr. Lisi can do his work on computers, and on paper, and seems to have various methods of support these days, but nobody would call them completely “mainstream”.

Lisi is a good example of how there might be ways of “doing science” that doesn’t require that one takes on a traditional trajectory in academia or industry. However, you’ll likely have to be dogged about it (Lisi’s been working at it for a decade), and pick a field that will be open to, and supported by, alternative ways of publishing ideas to your peers for pre-publication review.

An example of a mechanism for pre-publishing would be the site arXiv . When an author commits an article to arXiv, they are able to share papers and leave them open for searching. While it’s not peer-reviewed, it’s a good mechanism to expose your work to objective criticism while mountain biking at Wompy .

So, it was a pleasant surprise for me today to run into one of my former students at the Harvard Medical School lunchtime Systems Biology Theory Lecture, which is a nice tradition of a good lunch, a whiteboard talk (no powerpoint) and always interesting topics. He was there because he was working with the speaker of Friday (today’s) lecture.

The speaker was a mathematical biologist, Caroline Colijn who is studying mycobacterium tuberculosis. She is doing a double duty between the Broad Institute and the Harvard School of Public Health. Her talk was on flux balance analysis of TB, adding in information to the stochiometric matrix using expression data gotten from cultures of TB treated with treatment drugs.

Speaking of Broad, there is a good talk coming up on epigenomics given by Christoph Bock (Max-Planck-Institute for Informatics, Germany) on Wednesday Oct 31 at 10am in room 1001 in the 7CC Broad Institute building. It will be a lecture on EpiGraph (warning: small PDF from ISMB07). I would paste the entire talk abstract here, but it’s pretty long. Suffice today, the main point of the paper is that there is a “statistical epigenome” that needs to be represented, and they’re developing the tool to do it. Will probably prove to be very interesting.

The talk is by Dr. Randy Pausch in 1997 and I wish I had seen it when I was a grad student.

If you haven’t run across Dr. Pausch yet through email, or through the media or browsing Wikipedia yet, you may want to get to know him through his lectures and his projects. He’s helped a lot of people in his life, and by extension many more without realizing it was going to happen that way. He’s someone to listen to, if only that he’s naturally earnest and intelligent, and to paraphrase him in the Time Management talk, “experience is valuable”.

For those who want a shortcut, you can find his famous and fabulous CMU lecture here at Google Video

What follows is my own opinion on the issue based on my perceptions.

I want to point in particular to science faculty, but I suspect this applies across the board to anyone employed in the sciences.

Consider that self-identification and projection are probably significant obstacles to increasing faculty diversity

Increasing faculty diversity will likely require individual and institutional sensitivity for typical human bias towards what is usually termed ingroup identification, where individuals tend to exhibit favorable bias to those people they recognize as belonging to their peer group. This subject is difficult and will likely make many people uncomfortable. However, it is not necessarily the most negative feelings within our personal biases that may be hindering attaining reasonable faculty diversity, but rather more importantly positive bias towards those people whom offer us a quick social shortcut into identification.

An integral part of initiating social interaction (as I understand it) is the establishment of common ground—understanding of another. The projection of “self” onto another person greatly simplifies the process of social interaction as common ground can already be assumed. Communication will seem to go more smoothly. Trust is more easily established through a sense of familiarity, so collaboration might be easier. Some of this identification process likely requires social or class clues, but I’m just going to state the obvious and say that like most species, humans also base self-identification strongly on appearance which is then followed by other social cues.

I argue that self-identification is a limiting factor in compromised social interaction skills. The subtle point is that as professionals, we may not be conscious that we are socially discriminating against another individual if we are not feeling identifiably negative feelings about that person. However, without social identification and common ground, the effects of neutrality are likely as damaging as overt negative feelings could be. Neutrality would be acceptable if we could apply neutral social discrimination across the board and not favor some over others. As social animals, we will find perfect neutrality impossible.

Scientists will generally suffer from this kind of unconscious favoritism, I feel, even more than the average person. Although there are exceptions, in my experience, we scientists are not always known for deft socialization skills which would allow for ease in finding important common ground between us and those with diverse backgrounds. Since scientists make decisions on everything from grants to departmental resources and tenure awards on ‘best fit’, the dangers of social categorization and identification could be quite significant, especially if a significant part of any administrative decision is based on subjective perception.

To consider this issue with a real example, see the article by a former senior MIT faculty member, Dr. Frank Douglas that was also mentioned in Corie’s blog. Dr. Douglas makes the observation that the normal tenure process of discriminating for academic excellence is likely affected by more personal forms of discrimination.

Dr. Douglas decided to resign from MIT on June 3, 2007 as a result of his observation that critical issues for minority faculty at MIT were not likely to be addressed, in his estimation, by the Institute’s administration. Dr. Douglas decided he could not properly fit where an administration was unwilling to address the problems of discrimination in the academic environment, nor did he feel he could he advise young Black faculty as to how they could navigate the tenure process themselves.

Dr. Douglas does not question the tenure process itself, but rather the ability of the tenure process to carefully guard against the biases of personal discrimination.

If the success of the tenure process is partly reflected in an expected diversity of faculty, then one could argue some part of the tenure process may be broken for those departments that radically skew from the observed demographics of their field. Can discrimination for academic excellence be protected or separated from personal forms of discrimination? Might neutrally involving an outside group in the tenure evaluation process protect and educate all parties, from the Universities through the faculty to the applicant?

Examine and diversify science pedagogy

Graduate schools need to examine the pedagogy behind graduate science classes at the least. It has been known for quite some time that females and males have different learning styles.
All my graduate science and mathematics classes at the U. of Michigan were taught by men during my time as a graduate student, with one two-week exception.The more mathematical the science being taught, the less accessible the pedagogy seemed to make the subject to a non-linear or collaborative learner who desires synthesis and discussion with the equations. Some of the faculty were very good at presenting synthesis before calculation, but in most classes I found that I had to work harder than many of my male study-group colleagues because the material seemed to be better taught or designed for their overall learning style. I suggest we change the pedagogy for mathematical sciences to include concept before equations or organize additional resources for non-linear thinkers. It makes for thicker books and more effort on the part of instructors, but can yield better scientific intuition.

Is perceived attractiveness a factor?

Many people are socialized from an early age to assign social roles or worth to others based on how attractive they look. I’ve attended meetings where I’ve been the senior scientist in charge of a project but people were not aware of my role (though I was sitting near the head of a table), and as the conversation started I was addressed/attended to much differently than someone who was perceived to be more attractive. Developing a thick skin is important if you want to be a scientist, as is cultivating a healthy sense of humor, so when I notice that kind of odd behavior, it makes for a secret grin on my part and I move on. However, how do we address this? Female faculty at MIT brought evidence to their administration that they were treated differently than male scientists and MIT has conceded bias against women faculty. The same hazards of self-identification apply as above, probably deeply confounded with human socio-sexual roles. As for attractiveness, I’d hazard to say that most women are not perfectly in line with this culture’s ideal of “highly attractive”, and if some people unconsciously gauge how interesting a woman scientist is by her subjective reproductive worth, we’re not going very far with this science thing.

Why does industry attract more minorities and women?

I’ve known many talented minority and/or female scientists in my time. In fact, many of them work in industry. Why are they in industry and not in academia? It’s not as if industry is less demanding of talent than academia is. I’d say that industry is more demanding on many other levels, and from personal experience, the work can be as intense. The reward system, however, is much different and generally there are more minorities and women in the lower-levels of industry than in academia, though going up the ranks it becomes less diverse once again. Despite this, why are many women and minorities deliberately moving to and remaining in industry and not academia?

Change science culture

Since science culture is based on human values, asking science culture to change is probably as difficult as asking people to stop privately gauging attractiveness in their social interactions with others. However, the science culture in many of the physical sciences values independence and personality. The cult of personality and “big names” could be perceived as having greater value than collaboration and consensus. Projects might seem less important than the labs they’re coming out of. I’m not slamming big labs, here. Excellent scientists should continue to be valued and rewarded, and I’m not bashing those who have good names. I’m pro-achievement. I hope established scientists are encouraged to reach out and collaborate with smaller groups, especially those headed by minority and women scientists.

Also, this is not to say that women and minorities are innocent of playing the “big name” game, but we should encourage students to value collaboration and consensus as much as star status if we expect the scientific population to continue to become more diverse. There is the challenge that good collaborative work may not help distinguish a young faculty member during the tenure process, but I hope money continues to go to collaborative projects that have at least a small training grant for minority and women students/postdocs attached if they can’t get diverse people to join. Which brings me to…

Last, but certainly most important: mentoring
We should all attempt to mentor students and postdocs, and make that a priority. In graduate school, I ended up scrambling alone to figure out what I’d do for a postdoc position. My advisor was helpful but perhaps because of his own schedule could only offer minimal guidance. I’ve heard similar stories from other women scientists—mentorship, or the perception of mentorship—was something that seemed to be lacking in their experience. Departments should make (or continue) efforts to provide mentorship to graduate students (and postdocs) in addition to the mentoring offered by their advisor.

Early mentorship or collaborative projects could also give young scientists the ability to tune themselves into the science culture of their field, which might address several of the challenges in the paragraphs above, including allowing students and postdocs to tailor their own work towards obtaining a faculty position as well as becoming part of—and enriching—the culture of the field they want to enter professionally. MentorNet, a website that pairs underrepresented students with mentors drawn from the scientific community, has recently concluded a study that shows underrepresented students are more likely to value mentoring but may not feel they are getting the mentoring they want.

There’s a lot that needs to be done before we can see women and minorities moving into faculty positions. These were just some suggestions.

I’m not going to attend ISMB this year, unfortunately—a confluence of new job and busy schedule. Next year, ISMB will be closer to home in Toronto. I’m looking forward to it, because it will also give me a good reason to stop by and visit a scientist I know, Professor Laurence (Larry) Moran up at U of Toronto.

Dr. Moran is interested in many things, including the molecular-level effects of evolutionary processes. If you read the Usenet newsgroup talk.origins you’ll find some very good posts by Larry on many topics, including neutral selection. His website linked above has other interesting sites he’s authored, I recommend them. He also blogs at Sandwalk: strolling with a skeptical biochemist

Over the past few years, I’ve had some interesting discussions (sometimes heated) online with Larry on neutral selection, biological “noise”, and alternative splicing. For example, in the past discussions Larry’s viewpoint was that the majority of alternative splicing was noise with some functional exceptions. My viewpoint at the time, back in 2003 or so, was that there was a definite use for alternative splicing with specific examples, and we discussed the possibility that many of the alternative forms were in fact nonfunctional noise that yet provided a selectable background of protein forms for evolution.

Until recently, I had been working at a biopharma company that was interested in alternative splice products…like many other biopharmas at the time. The idea was that the discovery of an alternative splice product would be a chance to get some intellectual property rights on a protein and its use. I had seen countless examples of such alternative splicing products when combing through EST databases. Some genes had a plethora of alternatively spliced forms. In fact, we published a paper on the complexity of a particular GPCR family, the LGR receptors, in Molecular Human Reproduction in which we found several alternatively spliced variants of the receptor that seemed to have expressed protein products that had a functional activity in vivo.

There is other evidence for the complexity and functionality of splice variants, see for example this excellent open access review, Benjamin J. Blencowe Cell, Vol 126, 37-47, 14 July 2006 . The review is complex and detailed and I can only suggest you read it if you’re interested, as I can’t hope to do it justice here. One of its subsections gives a discussion of the global consequence of alternative splicing. Acknowledging the stochastic nature of splicing, and that many transcripts may in fact have no apparent biological function, the author proposes a new level of regulatory complexity: an alternative splicing “network” which may further regulate cellular proceses by providing different isoforms in different contexts—essentially, for specific interaction coordination in different tissues. He writes, “An emerging model is that these subsets of genes may comprise “layers” of gene networks that coordinate specific cellular functions.”

At ISMB/ECCE this year will be several talks specifically discussing alternative splicing, both in evolutionary terms (primates), in humans (ENCODE), and in specific cases of certain proteins.

One such presentation will be by Michael Tress, who was the first author of many on a PNAS paper detailing analysis of manually annotated splice variants in the GENCODE project (Tress et. al, The implications of alternative splicing in the ENCODE protein complement. PNAS 2007 Mar 27;104(13):5495-5500 ) where the authors examined a small segment of the human compliment of alternative splicing: about 2600 annotated transcripts for 487 distinct loci, with an average of 2.53 transcripts per locus. Supporting the papers mentioned in the Blencowe review, the Tress publication finds that in the loci they examined, that ”...these functional alternative isoforms appear to be the exception rather than the rule.” They also conclude that many isoforms may be deleterious based on detailed structural analysis of the protein products, but they note “If alternative transcripts in low numbers do not adversely affect the organism, the selection pressure against exon loss or substitution will be reduced, and the new variants will be tolerated, making large evolutionary changes possible.”

So, where is work on alternative splicing going to lead in the future? Blencowe’s review mentioned that there is a significant number of human SNPs that may cause disease phenotypes by affecting splicing. These diseases may result in aberrant splicing in the affected gene, causing loss of a transcript by nonsense-mediated RNA decay (NMD) or causing loss of a protein domain or protein interaction region, thereby disrupting the operating characteristics of the protein. There will likely be future work in discovering these SNPs and variations leading to alternative splicing. Additionally (also in Blencowe) there will likely be further work in examining patterns of alternative splicing between species, between tissues, and between specific developmental and biochemical contexts.

There is a lot of room for new research on alternative splicing. I recently attended a Gordon conference on Bioinformatics which I cannot discuss in detail as many results were not yet published. However, I can mention that I saw some good talks by Boston labs on alternative splicing: “Cooperative, Compensatory and Context Effects in Pre-mRNA Splicing” by Chris Burge at MIT , and “Polymorphic Splicing in Humans” by Hunter Fraser at the Broad Institute.

So, when I return to Toronto for ISMB 2008 , I plan on having a good visit with Larry, and discussing how he was right in that most of alternative splicing is probably noise. A conversation in his lab in Toronto a few years ago was on the right track: it appears most of sampled human alternative splicing may be selectable noise as there are functional splice variants among that noise that might allow interaction networks to fine-tune their behavior in subnetworks of functionality.

Stochastic processes that rule such effects as non-homologous recombination, alternative splicing or non-specific molecular interactions may be tolerated as they generate a selectable background for occasional evolutionary leaps. In the cases where suboptimal splice products may become dominant in a system for a gene with an essential role, a higher population of nonfunctional variants may lead to sub-optimized networks, which in humans yield phenotypes of disease.

So, when I read the CNN article entitled, Debate evolves into religious discussion I wasn’t surprised to find that three candidates for the Republican Presidential ticket are once again bashing one of the best supported scientific theories in favor of their own faith.

Throwing reason to the winds in favor of a gut feeling or faith might be occasionally viewed as personally irresponsible in a day-to-day setting, but if a religiously-motivated ‘gut feeling’ is held by a public figure with authority and power over armed forces, it could be dangerous.

Suppose a President had a gut feeling that we should go to war with another country with opposing religious views? Suppose that gut feeling was based on faith that he or she was guided by a higher power and shaky evidence rather than reason, logic, and lots of evidence? It seems rather apocalyptic—accepting faith over reason, and then hoping to govern with that philosophy. If your cynicism meter is going off, it’s calibrated correctly.

I’ve been following the anti-evolution political camp for a while. In 2003, I was at a Gordon Conference, sitting at a table with other attendees. The anti-evolution topic came up, along with a discussion of the “intelligent design” (ID) movement. I expressed my concern that the ID camp and creationism was something to be concerned about here in the U.S. The other people at the table pooh-poohed my concern, by holding that creationism and the ID movement are made of a bunch of crackpots that couldn’t possibly have any kind of long-term staying power. Reason, they said, would win out in the long run, and creationism and the ID movement would fall by the wayside. I’m afraid that just pooh-poohing this kind of thing isn’t working. We need more education, and we need it fast.

Maybe someone should mention that part of the U.S. economy relies on the tools given by evolutionary theory. Most of us in the molecular biology field know that we use mathematics based on evolutionary principles regularly in understanding biological systems and genomics. Drug discovery, in many ways, depends on evolutionary theory to supply the logical framework and tools around molecule and sequence analysis, as one example. Evolution isn’t “just” an incredibly supported explanation for an extensive collection of facts. It also defines a mathematical tool that allows us to group sequences into a logical order. The industry needs students ready to embrace this kind of science, not deny it.

So, how to educate the next generation in ways that allow them to see the consequences of evolution with their own eyes? How to introduce the new biology into primary or secondary schools? In fact, how do we educate teachers on this subject?

I’m teaching a graduate-level genomics course this semester, and I’m lucky enough to have a couple of high school teachers in my class. It occurred to me this semester that the material in my class—say, one lecture—could easily be adapted to several high school lesson plans. The evidence for evolution, and with it the reasoning behind it, could be presented to the student, in ways that show the obvious sequence-based evidence.

I think the community could develop a strong collection of lessons including sequence analysis and the basis of molecular evolutionary theory in ways that allow students to view the consequences of evolution with their own eyes.

The ‘new biology’ is at a stage where it’s ready to be taught to the high school level. We need teachers able and willing to teach genomics and sequence analysis, and we need computational resources available that will do the analysis on a server-side, so all the students would need are web browsers to do analysis.

There might be molecular-based or sequence-based biology initiatives out there, but I haven’t found many.

We need future scientists, and we need education of the broader public, so that anti-reason positions, like those found in the CNN article above are very, very rare. Molecular biology is a good place to start, and shouldn’t be a topic that’s left for college.

Imagine a site filled with other specialists in your general field, and being “holed up” with most of these people for nearly a week, during which time you are able to see presentations on research that hasn’t yet been published. The atmosphere is often informal, and if it’s a good Conference, it’s not very ‘clique-y’ either. And cheap—food, lodging, it’s all included for a very small price. Often, there are funds for students and young scientists as well to help defray or cover the costs of attending a Conference.

Conferences require an application for attendance, so make sure to apply sooner rather than later, many popular conferences can reach capacity very quickly.

Schedule of Gordon Conferences are usually in the same format: morning-to-lunch presentations, then a break in the afternoon for whatever (collaboration/discussion is encouraged) and then dinner. After dinner, a few more presentations and then a relaxed social hour (beer and chat) afterwards. You get to rub elbows with some great scientists, from grad students, to postdocs, research scientists and faculty. You’ll run into people you might consider legends that you wouldn’t normally expect a chance to meet in such informal surroundings. You come away eventually with a set of “Gordon Conference Stories”—often funny experiences or amazing conversations that are the start of something bigger.

More importantly, you’ll get some insight into some great new research (which you cannot discuss in public without permission as it is unpublished). Tradition says that the dinner on the last evening is a seafood (lobster) dinner, but don’t quote me on that.

Please, do yourself a favor if you haven’t looked yet, and visit The Gordon Conference Website where you’ll find the 2007 and 2008 list of conferences. The conference list changes every year as many conferences are on an every-other-year schedule.

There is also a special issue of Science magazine every year devoted to the Gordon Conferences.

There’s some down-sides too. The Conferences can keep costs down by hosting in schools/colleges. Sometimes the food is rather ordinary, and the sleeping arrangements—completely depending on the hosting site—might be a dorm bed. There might be shared bathrooms. There might not be air conditioning at every site. That’s not to say all sites are inconvenient. Some sites are fantastic, and there are usually local hotels for those who don’t like a certain site.

Also, many conferences are populated by people who have known one another for years. Don’t be afraid to talk to people, let them know who you are, introduce yourself around. If you plan on staying in the field that the Conference covers, these people will likely become some life-long colleagues.

I’ll guarantee you’ll find some of the best intellectual experiences you’ll ever have.

The trend in some of the big institutes like Broad has PIs in charge of large labs, postdocs and graduate students mixed with semi-permanent Masters-level technician staff to provide services from information management and data mining to lab maintenence and experimentation. The bigger the project, the more important the technical staff will become.

In bioinformatics, if you’ve been checking the job boards, you’ll notice a large number of computer programming and bioinformatics positions coming open in the Boston area, many at Broad and MIT followed closely by jobs in the Longwood area. Typically, their educational requirements are at or below the Masters degree level, following these employment trends.

Over the past five years, we’ve seen increasing trends for large-scale biological experiments and their associated information to become more standardized and less esoteric. Doctoral-level experience is no longer necessary to manage this kind of data or to conceive of interesting in silico experiments to perform on the data. Several of the graduates from the Bioinformatics Masters Degree Program at Brandeis are functioning in these “big science” groupsin exactly that way. Others are in large pharma companies supporting research and information management. Others are working in biology cores in small groups in the Longwood area.

Molecular biology jobs at the Masters degree level do not seem as active on the job boards right now. I believe this will probably change with the continual standardization of experimental procedures and the growth of standard protocols for in the “big science” models. There is a definite need of chemists and biologists who can adequately design and carry out experiments who nevertheless do not need the training of a PhD to run, say, a bench within a sequencing core or a chemical library database. I don’t pretend to be a fortune teller and you’re right if you think this “prediction” is obvious, but it appears that as biology gets bigger and more centralized, chemists and biologists with Masters degrees will find themselves more in demand around the Boston area.

Biostatistics is still pretty much a doctoral-level type of job at this point. This is for several reasons that spiderweb throughout that employment market, including the necessity of trained people in clinical trials and some of the difficulties in addressing standard types of educational needs at the Masters level. However, I predict that there will be a growing need for trained Masters-level medical informatics and clinical informatics technicians within the next few years. As the number of clinical trials grow, so will the need to tie that information back into drug discovery projects, and the federation of existing bioinformatics and chemical databases with medical and clinical trial data will be necessary. It’s on the horizon now.

Another lesser-known area that I believe will grow and diversify is the “chemoinformatics” area. Although this field is currently closely tied with structural biology, I believe that there is room for Masters-level trained chemists-slash-programmers to combine their education with experience in structural biology to assist large-scale screening projects as big projects seek to explore chemical space in everything from cancer to parasite biology.

Today, the Guardian UK gives the headline, “Expert casts doubt on flu vaccine” in which the first line of the article is, “Flu vaccines may not be as effective as people think, an expert has warned.”

Responsible science reporting should not sum up a complex article on the need for better epidemiology studies as “flu vaccines might not be as effective as people think” because one can interpret that statement in many different ways. In a fast read, it might seem that flu vaccines are being called ineffective.

The background: in the British Medical Journal, Tom Jefferson, co-ordinator of the vaccines field of the Cochrane Collaboration, stated that there was an urgent need for re-evaluation of vaccination campaigns for safety and effectiveness precisely because the data and the problem is so complex.

According to parts of the Guardian, however, it appears that Mr. Jefferson is calling into question the effectiveness of the vaccine itself.

I’ve looked at the original article at the British Medical Journal and found that the author is painting a complex picture, stating that the health picture of influenza is complex, has limited value for public health policy and safety studies, and needs to be re-evaluated.

Mr. Jeffersom provides the numbers, and the reasoning behind his evaluation.

It’s a good read, but it’s not saying flu vaccine isn’t effective. Mr. Jefferson is saying that there’s little evidence that the vaccine is effective in particular groups under particular measures because of difficulties doing the studies in question.

A lack of evidence because of difficult studies isn’t “may not be as effective”. Someone should tell the Guardian that. I’ll wait to see how many other news media jump on the “flu vaccine not effective” bandwagon.

See article here: http://www.guardian.co.uk/uklatest/story/0,,-6174061,00.html

See original BMJ here: http://bmj.bmjjournals.com/cgi/content/short/333/7574/912?ehom=#TBL2

I find that a difficulty arises because the terms associated with information theory are often loosely conflated with general biological processes, often above the individual molecule level. For instance, gene processing from DNA to protein in the “central dogma” of biology is often termed as an information flow.

The term “information flow” begs a deeper question: what, if anything, can be measured as ” biological information?” What sends that information, what receives the information, and what is the process that allows information to “flow”?

I like how Dr. Thomas D. Schneider (at NIH) defines information:

“Information is always a measure of the decrease of uncertainty at a receiver (or molecular machine).”

Using this definition, the concepts of sender and receiver are critical to understanding the message being passed.

The term “information flow” in biology implies an implicit understanding of a process between a sender and a receiver, and the nature of the message being passed. The idea of “sender” and “receiver” are difficult to map onto a biological system, and there will be very few instances where this could be cleanly applied.

One case of a clearly discernable message might be the molecular interaction of the MHC-antigen system with the T-cell receptor. It’s a fairly clean system to postulate “information flow” at the level of hydrogen bonds and molecular recognition. The players in the process are clearly defined. The information is presented and received by specific mechanisms. However, the ultimate sender and receiver of the information aren’t necessarily the receptors themselves, as the actual information needs to propagate to the immune response—in a sense, the receiver of that specific molecular information is actually another molecular process. In larger systems, or systems that cross multiple scales, the receiver of the information may be blurred in its distinction between the roles of sender, message, and receiver.

Without a good definition of a receiver, we cannot understand the character of the message being transmitted. Without a technical understanding of a message, we lack a definition of the actual information, and (obviously) without a definition of the information, there’s no “flow” we can define with any applicable certainty.

In the abstract, “biological information flow” sounds good to the ear. I’ll stop short of saying that it’s meaningless, but unless the term is applied to specific systems under specific technical conditions, I don’t see its applicability to general cases.

Dr. Scheinder’s primer on information theory for biology can be found here:

http://www.ccrnp.ncifcrf.gov/~toms/paper/primer/primer.pdf

Promoters, those regions upstream of a gene, can control how much of a gene is expressed (“gene dosage”) and under what conditions that gene is expressed, all of which can vary from cell to cell.

The summed effects of polymorphisms, including those found in promoters, may be important factors at the heart of predisposing conditions for complex diseases.

Recently, papers have been published which implicate promoter polymorphisms for several complex and acute diseases, including RA, diabetes, sepsis, atherosclerosis, and depression.

On the 19th of October 2006, a group at Vanderbilt (Campbell et al) have found a p=0.0005 association with promoter variants of MET receptor tyrosine kinase, which significance rises to p=0.000007 in families with more than one autistic child. The data implicate reduced MET gene expression in autism. The researchers pegged MET partly by their observations that MET might be associated with the observed gastrointestinal and immune symptoms that are often reported to accompany autism.

It is significant for its direct genetic link to autism across multiple families.

The polymorphism is also interesting in that it seems to confer a 2x decreased MET promoter activity.

Citation:

Daniel B. Campbell, James S. Sutcliffe, Philip J. Ebert, Roberto Militerni, Carmela Bravaccio, Simona Trillo, Maurizio Elia, Cindy Schneider, Raun Melmed, Roberto Sacco, Antonio M. Persico, and Pat Levitt. “A genetic variant that disrupts MET transcription is associated with autism”. PNAS published October 19, 2006.

http://www.pnas.org/cgi/content/abstract/0605296103v1

But.

Recall this past decade, when Bioinformatics was getting the same kinds of rush-rush-buzz-promise stuff that we’re hearing now about Systems Biology.

I’m not saying we were naive in the 1990’s and early 2000’s, because I think even then we were trying to be realistic. We didn’t think the genome was going to necessarily yield the grail of pharmacogenomics, any more than we expected microarrays would eventually yield perfectly predictable cause-and-effect at the transcriptional level. Biology once again was revealed as complex and recalcitrant to simple models. Many scientists in industry and academia continued to represent the information-driven paradigm as the next wave that would make medicines cheaper, make drug discovery faster—and give whole new therapies that nobody even dreamed of.

We know what happened when the promises didn’t deliver. Granted, drug discovery is a slow process, but computational biology and bioinformatics are generally given a wary eye outside of academia. They didn’t deliver fast enough, so therefore this kind of discovery-driven approach was viewed as a failure.

This mistaken perception should give us a lesson, now, in systems biology. We don’t want to lose the faith of the decision-makers, because systems-level biology research is absolutely necessary for advancement in medical science.

Bioinformatics and computational biology are necessary parts of the effort. The most successful projects in computational biology tend to couple tightly with biology. Similarly, systems biology requires full support from bioinformatics and computational biology, and this should be clear from the start.

If you haven’t read my earlier posts, I’ll let you know that part of my concern is that systems biology, as a paradigm for doing science, is going to inherit the same challenges still facing bioinformatics and computational biology, but further compounded by more challenges in complexity.

In the bioinformatics community, we’ve tried to generate standard protocols for interpreting problems in tech like microarray analysis (for example) and have generally not succeeded. The protocols are there, and there are some good methods for data analysis, but everyone has a different favorite experimental platform and post-processing method. Experimental conditions are not standardized even on the same cell types. All this (and more) confounds federation of experimental results. The difficulties in generating consistent results between experiments aren’t a failure of bioinformatics or mathematics per se. The inconsistencies, outside of a lack of standards, are a consequence of biology—more specifically, challenges in chemistry and molecular-level physics. In fact, most of the problems with bioinformatics analysis of data sets arise from the reality of the data sets. Biology at the molecular level is not usually repeatedly, predictably deterministic. Models are built to take noise, and these molecular processes, into account.

Additionally, high-throughput methods, as they get miniaturized, get more prone to statistical noise at the molecular level, and therefore more prone to variation.

Systems biology is challenged by these facts. Molecular science is heavily stochastic. Molecular biology is highly variable even between “identical” cells. Systems biology is by definition, inherently molecular. We all face the challenges of understanding biological systems, which are controlled by multiple parameters that need to be modeled within multi-component, complex time-dependent multi-hierarchical systems governed by stochastic phenomena at all levels.

Given all these challenges, I fidget a bit when I hear that systems biology’s goal is better and less expensive clinical trials. Our goals should be more immediate: let’s do a good job at figuring out the cellular consequence of insulin signaling, for instance. I don’t want to be the voice of doom, but I also don’t want industry or government giving up on systems biology approaches when they find out it’s even harder than basic bioinformatics problems were. In fact, we haven’t even gotten to a satisfactory place in a lot of bioinformatics and computational biology challenges, including protocols, communication and data federation. As a community, we’re still addressing those questions with projects like Bioconductor for common protocols in statistical data analysis, and BioPax and SBML for common data exchange formats.

Let’s also remind people outside the direct scientific community that we generally don’t know how most biology works at the molecular level. That’s the point of systems biology, after all—to try to figure that out. We might have to start simple. Part of being responsible is saying that we don’t know all that much about what we’re doing. Let’s not let the promises be made for us by keeping quiet about the scope of the problem.

What should we try to do, right now, in order to make sure that systems biology isn’t written off as “too hard”, too soon? How can we avoid systems biology being prematurely labeled as a failed approach?

Here’s my suggestions, please add your own in the comments. This is a community effort.

1) Encourage responsibility in public speaking and in industry partnerships. Academia should be honest with industry as to what’s possible using systems approaches. Small companies should try to refrain from promising to solve all of pharma’s problems with systems biology. Industry and investors should be wary of too many concrete promises in such a new field. As a community, we should encourage industry and academic leaders to speak to the challenges ahead, as much as to the promises. We should remind those who are interested that results won’t happen tomorrow. We have to start working on this problem now before we can expect to start getting results. IN fact, we should expect that most great initial results will probably be as much serendipity as good design and skill.

2) Start in a focused manner. If we can’t start small, we should start focused and not spread our whole community out over many different projects when we might succeed in focusing at least a few large groups on a single project.

3) We have to pay attention, close attention, to education. The International Society for Computational Biology, for instance, has been mulling over the challenges of good computational biology education. We should be more proactive in training biologists to be computer-literate. If a University is going to start a Systems Biology group, they should include a biology degree that will enable a student to function in that group.

4) Actively re-train physical scientists. Don’t make computational biology or systems biology inaccessible to adults who are already trained in one discipline. Encourage them to re-train and come on board. I recall a story of a highly qualified physics PhD interviewing for a job in a systems biology group in a large institution here in the Boston area, being embarrassed when he accidentally heard someone asking a hiring manager why they were bothering to interview an “adult” instead of a “kid”, for a key scientist position in their modeling core? For me, this is confusing. Systems biology projects need, most of all,the sum total of our experience.

5)As we had the Human Genome Project, we need a Human Systems Project. The NIGMS is forming national centers for Systems Biology. We need to move on something wider and multi-centered with set protocols and common language. Don’t close it down to the privileged institutions. Open the effort to all Universities around the world by making data available from a National Systems Project where protocols, cell types, and systems can be better defined.

It would be great to see some more serious dialog out there, instead of rushing to see who can get the most money the fastest with the most faculty on board. We’ve got to make sure that we don’t make people bitter with promises of too much, too soon. If Systems Biology is going to transform biology as we know it, it can’t lose the confidence of the decision-makers.

I am mildly concerned that there’s an inherent problem in a pell-mell rush towards systems biology. I’m concerned that there are no efforts to educate biologists capable of working in complex systems at the undergraduate level, or even at the graduate level, any more than we really, as a wide community, tried to widely disseminate computational training in biology. Systems biology extends that need. How many undergraduate biology degrees require extensive computer experience, physics, engineering or complex systems courses? Are we going to re-assemble the pieces of “old science” and call it “systems biology”, or are we going to attempt to educate the next generation of systems biologists?

We’ve got to re-train many of our existing biologists to be literate in the computational side of systems biology—not necessarily as expert programmers, but as biologists who can understand and work within the demands of the field. Without involved, literate biologists with a combination of lab and computational skills, “system biology” is going to be a paradigm that doesn’t differ significantly, at the scientist level, from what we already have.

We also need biology-literate classically-trained technologists, engineers, physicists, and chemists. What are we, as a systems biology enterprise, doing to widely encourage and develop this kind of education?

The community should make it a point to say right off, to anyone who cares to listen: some people are going to make promises about systems biology that will be hard promises to deliver on. I’ve heard it said more than once, in symposia and in discussion, that systems biology might make drug discovery more successful at either the discovery end or by helping to make clinical trials more successful.

Those kinds of statements will perk up the ears of biotech and pharma, who spend millions of dollars in failed clinical tests. Failures from from toxic side-effects, and also from a failure to show efficacy. Sometimes, a failure to reach a trial’s designed output might not say anything about the drug’s success in patients, but rather in conditions of the trial, such as the company’s choice of patient to test the drug in. If there is a way to increase success of clinical trials, even by a few percent, that means cost savings of millions of dollars. It’s suggested that systems biology approaches can work towards improving clinical trial success rates. Pharmas, biotechs, and large Universities are well aware of the potential of systems biology in this regard.

Promises of more successful clinical trials might be fulfilled someday but possibly with a caveat: I suspect that increased clinical trial success will come at a price I haven’t heard discussed. Along with the systems approach to translational medicine and use of this data in clinical trials, there is the inevitable need to segregate subgroups of patients in smaller groups, because systems biology will discover the diversity of phenotypes on genotypes and each type will require testing. We may have more successful trials, but I suspect they will be successful trials of smaller populations.

Expense may also go up. We’re likely going to see two additional sources of cost at least in the shorter term. First, the need for many more data points per subject to support a systems approach during a trial or preparing for a trial, and second, a greater multiplicity of stratified groups needing to be tested to ultimately make a good-sized target population for a particular drug’s market. Smaller populations successfully targted with a fully developed drug might mean more expensive drugs unless we can target multiple populations—which might mean multiple trials. I think any group will be lucky if it can find clear cases where systems biology approaches will yield information on a large biallelic population that easily segregates out for a clinical trial.

So, as planned, we can apply systems biology to clinical trials, or to drug discovery, or to translational medicine—and we might get lucky in some cases. But, as people in the field for the last few decades already know, biological systems are difficult to model for the fact that the data models aren’t out there, yet. The more detailed the model, the more data needed. There are sophisticated computational models that can model limited or simple systems right now. I’m impressed when I can see some accurate predictions on carefully controlled behavior in prokaryotes or a nice prediction on a biomarker set in a translational oncology setting. There’s good work out there, and I’ll be the first to shove a paper at you about some cool study or another. I’m no pessimist about systems biology as a paradigm for biology. I’m fascinated by it.

While visiting California for an educational workshop on computational biology, I mentioned to Doug Brutlag that I was starting to teach a systems biology course, in a 3-hour-a-week, 10 week format. He thought I was a little crazy. Systems Biology, he told me, was probably too huge a subject to distill down into one semester, let alone a 10-week night-school format. Doug was right—we could probably do better than a 10-week survey course, but it’s a necessary introduction into a field that’s growing by the week, let alone the year. I think the course works as well as it could. The problem facing me now, as I get ready to teach it next semester, is in keeping up not only with what is advancing in the field, but also keeping up with what people define as systems biology.

The field of computational biology has evolved quickly over the past five years. I’m now the Program Director in the Bioinformatics Program, and here at Brandeis, we will undoubtedly face the challenge of developing our bioinformatics curriculum to be more geared towards modeling and systems biology. We want one of our objectives to be re-training biologists and other scientists on how to be computationally literate—in a range from sequence analysis to flux balance analysis.

It’s hard to say where to draw the line, though, for what’s important to systems biology, when it seems to encompass, at this point, everything under the sun.

While it once seemed silly to try to put systems biology together in a 30-hour semester, these days, it seems nearly impossible not to get a lot of systems biology buzz at every conference. Systems biology can now be summed up into the “Four M’s” buzzwords on single slides.

It’s getting so that it’s impossible to keep up with the rush towards systems biology around Boston these days. An acquaintance of mine at Brandeis noted that he had recently spent three separate meetings with three different groups of people—from three different consortia on systems biology in the last month alone. Whole departments are being set up in industry and academia based around this “systems biology” paradigm. While I think it’s a great idea, it seems to me like anyone who’s analyzing high-throughput data these days has stopped using the words “computational biology” or “bioinformatics” and has just started coupling the informatics with experiment and renaming the whole enterprise “systems biology”. That’s not what I’m concerned about, though. I’m concerned about the backlash that might occur when the first flush fades away.

I’m a little suspicious, always, of the “flavor of the month”. Systems biology seems to be the flavor of the month in the Boston area, at least. It feels too much like the big rush towards bioinformatics ten years ago. Bioinformatics will soon be dead if you believe Lincoln Stein. Maybe bioinformatics will be dead because everyone will start re-labeling all of its work as “systems biology”, but I don’t see any danger in computational biology or bioinformatics vanishing, no matter how you define them.

Work in systems-level biology isn’t new. In fact, there have been many good projects active in the field for years, headed by people like Hiroaki Kitano, Masaru Tomita, Eberhard Voit, Andreas Wagner, Michael Savageau, Bernhard Palsson, George Church, Trey Ideker, Leroy Hood — and so many others I’m leaving out and saving for another post. For those of you just starting in systems biology, those names are a good place to start your reading if you’re interested in the field.

They’re also good people to learn from. Scientists and decision-makers should look carefully at the challenges faced by those groups over the past decades and then learn from them. Systems Biology can deliver interesting science, but it’s a slower type of delivery compared to what might be expected in the shorter term. Systems Biology is going to be a long-term process. Our expectations should match up with the experience of seasoned systems biology scientists.