Saturday, April 28, 2007

Degrees of Difference

The recent abrupt departure of the MIT admissions dean provokes many thoughts, some silly ("Not the type of admissions she was supposed to make!") to the serious. But one thing which it certainly underlines is the limits of degrees as indicators of ability.

I by no means condone Marilee Jones' resume stretching. I remember near thermonuclear conditions when once someone nonchalantly described a third party's intent to obtain an advanced degree (I think it was a doctorate) through a mail order college. I worked hard for that sheepskin! I sacrificed! I suffered! Thou shalt not claim degrees in vain!

But, degrees have at least two relevant properties. First, they are a symbol of hard work and achievement. Second, they are proxies for ability and knowledge when attempting to hire someone. Ms. Jones never did the work or made the achievement, but she certainly demonstrated the aptitude to carry out her position. She was apparently widely recognized as a leader in her field.

It is less common than it once was, but some companies still require certain degrees for certain positions. In some cases this is completely defensible (could you take seriously a medical director lacking a medical degree?) and perhaps even legally mandated (is there any position that legally requires a Ph.D.?), but too often this is the lazy way out of real thinking. Degrees are important, but the level of effort invested varies. There are many talented persons who are perfectly capable of earning an advanced degree, but for one circumstance or another never did -- and some of them have gone on to Nobel Prizes. On the flip side, there are persons who no longer exhibit the qualities of a Ph.D. or M.D. -- persons who make you wonder how they got there in the first place. Even persons brilliant in one area can be completely off the rails in many others -- e.g. William Shockley.

I believe that one mark of an intelligent organization is an ability to look beyond paper qualifications and look at real achievement. My previous company had an ongoing tradition of this. When the big sequencing center was being set up, they didn't demand a Ph.D. but rather handed the job to someone who had shown previously (in academia) the ability to get things done -- he did & went on to found a successful genomics company (Orion). Even near the end of my tenure there were research associates being promoted to Ph.D. level titles ("Scientist"), because they had demonstrated the creative thinking and scientific leadership which are the true requirements of that grade, not the piece of paper.

Hiring people is hard & sucks up lots of time. Degrees are a useful shortcut, but one should never forget that shortcuts aren't always the best idea.

Thursday, April 26, 2007

(Still)Birth of a Neologism?

A proposal has been published in the open access journal Molecular Systems Biology for a new term, or really family of terms, for various elements of genetic information.

First they propose a whole host of types of genes. For example, a protein coding gene is a P-gene and these are further subdivided into structural protein genes (sP-genes) and regulatory protein genes (rP-genes). A cynic might point out there are already proteins refusing to choose sides, such as transcription factors with enzymatic activities (I know I've come across them, but alas memory is failing to return their names). Actually, they already lump the proteasome subunits into the P-gene class (subclass 4). Are indirect regulators of transcription (e.g. IkB and IKK, which regulate the transcription factor NFkB) in here too?

RNAs in this taxonomy come in two basic flavors: structural (sR-genes) and regulatory (cR-genes) c=control? cR-genes are subdivided into discriminating (regulating specific genetic subprograms; e.g. miRNAs or XIST) and non-discriminating (broadly acting; e.g. tRNAs and snoRNAs).

There's more. For all of the cis-acting elements controlling a gene are its 'genon' and the trans acting factors the 'transgenon'. We also have pre-genons, holo-genons, proto-genons, holo-transgenons.

In any complicated endeavor jargon is inevitable, as complex topics can't be explained in detail every time you go to talk about them -- rather, the jargon serves as a shorthand to enable actually getting something done. Attempting to generate such taxonomies is useful, but it's hard to think of much success in that department. This exercise is reminiscent of Brosius & Gould's attempt to create a nomenclature for pseudogenes. Very clever, but it never caught on.

A cynical pedant might be inclined to ask "what's the point of inventing new jargon when nobody can be bothered to properly use the old jargon". For example, periodically the popular press (and sometimes new iMedia of blogs) trumpet the discovery of a new human gene, which might be a bit disconcerting to various taxpayers who thought they had paid to have them all found already. While there are almost certainly some new genes to be found, in most cases what is new is an association of alleles with disease, and in these SNP-saturated times even the alleles can't claim to be new. Too often also is the overuse of 'gene' when the more specific 'locus' would do, or gene where 'gene product' or 'protein' would be a better fit. And other times, not only was there no discovery of a new gene, but the phenotypic association was linked only to a large stretch of DNA.

One speculation this all leads to is what does drive the acceptance or non-acceptance of new terms, particularly ones intended to be pronouncable (nobody, other than the once extant company, tries to read siRNA as two syllables!). 'omes and 'somes seem to have a better bet than some terms, but I'm sure there's been duds for that (and of course, words that didn't enter the language by that route -- do I really live in the collection of all things beginning with the letter 'h'?). Some good terms rise & fall, or only survive through their derivatives. Virtually nobody talks about a cistron, but polycistronic survives -- a pity, since cistron is a perfectly wieldy word -- luckily I can stay gruntled without solving the mystery.

Gobble Gobble Slurp

AstraZeneca's record-setting $15B+ buy of Medimmune gave the old workplace's stock a mild goose, but things have settled. It is a reminder of what the ultimate fate of virtually any semi-successful biotech company will be.

In the end, there are three possible fates for a biotech: survival, liquidation or acquisition. Liquidation is rare & will probably always happen to early-stage companies, but does happen. One genomics company (Progenitor) reputedly let their employees show up for work to locked doors. Most companies will be acquired down the road; only a few frontrunners will stay independent. There are, of course, variations on these themes. J&J has a track record of acquiring companies but then leaving them largely recognizable. Some companies (e.g. Cadus) disappear in an operational sense but never quite disappear legally -- business zombies. Mergers of equals are theoretically possible and often claimed (Biogen-Idec), but how lopsided the division of spoils is can't really be assessed by an outsider.

Millennium executed a number of acquisitions during my tenure, with many being quite successful -- influential people remain who joined through the Chemgenics or Leukosite acquisition. Leukosite brought in Velcade (then PS-341), from a company (Proscript) which Leukosite hadn't finished digesting (er, assimilating) when the the MLNM-LKST merger was announced. Much of Millennium's inflammation pipeline has strong roots back to Leukosite.

But then there was the big demonstration of 2+2<<4: COR. Millennium bought COR for Integrillin & a sales force, with some interesting early stage oncology and cardiovascular programs in as icing. The corporate cultures seemed compatible and the excitement was there. But somehow things quickly ran downhill & when it became apparent that Millennium was overstretched, the COR (now MLNM San Francisco) site was targeted for liquidation. Eventually, after sinking many dineros into further clinical studies, Millennium essentially walked away from Integrillin. So for $2B plus, a sales force was purchased plus a revenue stream from Integrillin and some other leftovers -- plus some important contributions from the ex-COR folks in wrapping up the Bayer collaboration. Was it worth it? My impression is that everybody on the COR side wished they could get a do-over.

MedImmune was hardly an isolated purchase -- big pharmas and even big biotechs (Amgen, Genentech) have been plucking out various biotechs, generally either for hot therapeutic platforms (siRNAs, advanced antibody technologies or exotic antibody alternatives) or late stage compounds. Looking around the Cambridge neighborhoods finds plenty of companies in the former (Dyax, Alnylam, Archemix) or latter (MLNM, Vertex, Alkermes) categories. The majors still have their gaping pipeline gaps, and Wall Street is starting to hound Genentech towards more acquisitions -- and Amgen is starting to experience the pain of commercial reversals. Odds are there will be more buyouts -- and more flameouts & companies (e.g. Imclone) which flop at the auction bay.

So grab a ringside seat & get comfortable -- but please don't play the ponies. If anyone tells you they know who's going to buy whom for what price, odds are they're lying. Even if they aren't, do you really want to follow in the footsteps of the famed biotech investor who was wisked from her Connecticut home to a federally-paid stay in West Virginia?

Monday, April 23, 2007

A good RNAi guide

Cell Cycle is an interesting little journal that publishes many papers open access. A nice little review of the statistical treatment of genome-wide RNAi is available freely.

The review focuses on noise and variance in RNAi screens, and doesn't explore some of the other key issues such as off-target effects, interferon response and appropriate cell lines. So it isn't a complete guide but a sharply focused one.

A recent Nature has a paper on genome-wide RNAi for targets increasing sensitivity to the key antitumor agent paclitaxel (alas, not free).

Genome-wide RNAi with siRNAs is a powerful technology, but it requires a pretty large investment in automation to make it work. That will slow the widespread adoption of the technology, which isn't entirely bad. In some ways, mRNA microarray technology spread too far too fast leading to many bad papers being published before the methodologies were well worked out. Of course, there are still lots of bad microarray papers being published, but you can't make the horse drink. Some bad papers have poor microarray analysis, and others are just atrocious experimental design. In the end, the technology has been besmirched, generally unfairly.

Nobel Betting

ChemBark proposes odds for next fall's Chemistry Nobel. Of course, a lot are really biology (prompting the usual grumbling from a small set of chemists). I would too argue along with one commenter that 'The Pill' would be more apt under Medicine (and would seem to be deserving given the impact on society).

One of those satisfying "I've really joined the fraternity" moments is when you start recognizing the names in the Nobel announcements as work you are familiar with. Of course, that doesn't mean it happens every October, but often enough that I'm still convinced I earned my stripes.

Friday, April 13, 2007

Farewell Kilgore Trout

The passing this week of Kurt Vonnegut was strongly felt here, as he is one of my favorite authors. I first was keyed into Vonnegut by a high school English teacher (by the name of French!) and started reading a few. When I got to college, I started plowing through the whole shelf of Vonnegut novels & catching his few last ones as they came out. If you've never savored the ever-relevant bitter humor of Mother Night or the absurdity of Cat's Cradle, you're missing out. Breakfast of Champions & Sirens of Titan are great fun too, and even the lesser works have lots of good bits in them. Vonnegut isn't for prudes, though has texts are really pretty tame compared to a lot of other literature before and after (e.g., my current soul-enriching time sink, The Good Soldier Svejk -- now there is some inspired vulgarisms!)

Please don't take this the wrong way, but I enjoy reading good obituaries. Not that I'm happy to see someone pass, but good obituaries teach you something you didn't know -- perhaps because you never knew the person existed (but should have), but other times because you didn't know something interesting about a familiar figure. I never knew that Vonnegut had studied to be a biochemist. I think Asimov was also originally a biochemist as well -- but surely to group them in this way is to engage a granfalloon. Hi Ho!

Thursday, April 12, 2007

When a little fighting is good

Last week's Nature has a great paper in it that takes several readings before it makes sense. It's not that the paper is poorly written, it's just that the results take a while to burrow through a bunch of preconceptions.

The paper looks at the effects of combining two antibiotics which antagonize each others actions, primarily doxycycline and ciprofloxacin. What is interesting, and initially counter-intuitive, is that mixing the two antibiotics at low doses (all in culture) creates a very different selective environment than two antibiotics which show additive effects. If you look at the space of possible combinations, then if the antibiotics add a strain resistant to one of the antibiotics will always grow better than an otherwise identical strain which is sensitive to that antibiotic. However, if the antibiotics are antagonistic then there will be some combinations which favor growth (Figure 2). They go on to show that this theoretical analysis indeed holds true for mixed cultures (Figure 3).

The interesting possibility is that by dosing the two drugs appropriately, resistance mutations would be at a competitive disadvantage to their wild-type kin and would therefore not take over the population. The article is appropriately cautious in extrapolating these effects to clinical settings. This remains to be demonstrated in an animal setting, let alone a patient. But it is intriguing. In addition to some mouse experiments, an obvious next step would be a further exploration of whether resistance mutations do emerge in long-term cultures.

Even if the results were to get past such experiments, it is difficult to see this work escaping the economic trap which ensnares all antibiotic work. Any commercial activities must compete with cheap, highly effective existing antibiotics. Academics could try to develop such combinations, but the patients in true need of such drugs are desparately ill, making finding any clinical signal difficult.

However, this could be a very interesting strategy to explore for cancer. There are definitely examples of chemotherapy agents which conflict with each other. For example, topoisomerase inhibitors & bortezomib. Combinations using this strategy might be cytostatic -- aiming to keep the tumor in check -- but that could be sufficient in many settings. However, these bump against the challenge for any cytostatic chemotherapeutic agent: you can't use tumor shrinkage as a clinical measure but must measure survival. Survival is what really counts, but requires much longer and larger studies.

Wednesday, April 11, 2007

The ups-and-downs of out-licensing

Monday's Boston Globe had an interesting article (probably $$$) on a story which hadn't seen much attention previously but illustrated a number of biotech themes: rapid reversals & odd partnerships.

A group at Beth Israel Deaconess Medical Center (BIDMC) had developed a potential new protein therapeutic which they thought they might hit it big with. I must confess a special fondness of BIDMC, as a team there oversaw the delivery of my most important project ever, but they seem to have really gone out on a strange limb in this case by picking an odd partner for developing their blockbuster.

The potential blockbuster is apparently a single chain protein encoding a dimeric erythropoeitin (EPO). EPO is, of course, the most financially successful biotech drug ever and what made Amgen bigger in market cap than some old-line pharmaceutical companies. EPO has been in the crosshairs of a number of other companies, but Amgen has thus far won all the battles on patents -- first with Genetics Institute (now Wyeth) at the outset and later knocking out Transkaryotic's (now Shire) attempt to end-run their patents. Amgen followed up with a slightly modified form (Aranesp), which has also cleaned up. Affymax has a clever mimic in trials -- though this illustrates the need for patience in this business, as their splashy paper on it came out when I was interviewing at Millennium nearly 11 years ago!

So a tandem EPO would seem like a reasonable bet, with the claim that this form is longer lasting (ala Aranesp) and more potent. EPO is used to treat anemia in kidney failure (EPO is normally made in the kidney) and cancer patients (along with illicit off-label uses in the field of athletics) -- more is better, right?

Lately, the bloom is off that rose -- several studies are suggesting that for cancer patients there may be drawbacks to high EPO doses. EPO has now earned a black box warning and Amgen is scrambling (the CFO just bailed -- perhaps with shoeprints on his backside).

Now that's just bad luck -- pharmaceuticals are like that. One day COX2 inhibitors are miracle drugs; the next day they're persona non grata.

It's the other half of the story that I found very curious. If you were trying to out-license your institutions exciting new protein therapeutic, what would your first choice of company be? How about a failing genomics company with a slim bank account? No? That doesn't sound appealing? But that's exactly what BIDMC did.

Now a lot of genomics companies exited the genomics boom in a strange place: lots of money raised during the bubble, but no path forward to make money in genomics. Companies such as HGS and MLNM are still living off that cash, but they had interesting programs going. Others had stranger outcomes. Variagenics and Hyseq proved that 1 genomics company + 1 genomics company = 0 genomics companies, as they merged, ditched all their genomics operations, and changed to Nuvelo to develop an in-licensed protein therapeutic.

BIDMC chose DNA Print Genomics for their wonder drug. I have nothing against DNA Print, but there's nothing in memory (or on their website) to suggest that they have any of the key skills. Nor do they have much cash. While they do actually have products, those products don't inspire much awe. DNA Print will type your DNA to estimate your ancestry. That might be fun, but how big is the market really? They also claim to have tools which can predict the physical characteristics of a person (skin color, earlobe attachment) from forensic samples -- a sort of genetic sketch artist. I'm sure there are missing persons-type cases where this provides one more set of clues, but its hardly something that would see routine use in cases.

Wall Street hardly loves DNA Print -- if Yahoo's statistics are to be believed, it is trading at a market cap of 4.58M much below its cash position of about 8.5M -- but it is also (if I'm reading this right; I really don't stare at these often) blowing through 3-4M per quarter -- meaning that cash will run out in the near future unless they find financing or take a scythe to their operations. This was a point raised in the Globe article -- BIDMC has hitched their wagon to a lame horse which may expire very soon.

An interesting question, which one can never get a straight answer to, is why pick DNAPrint? Was there really nobody else interested? It is curious that the consultant who BIDMC hired to find a licensee (after Eli Lilly had bailed out) for the compound ended up as chief executive at DNAPrint. While that is hardly unheard of, it does raise a real issue of conflicts of interest. The deal structure is strongly loaded towards milestones & royalty payments -- i.e. BIDMC sees very little without a lot of progress being made. DNA Print apparently has reported preclinical results.

A weak & failing partner for a troubled market niche -- hardly a good place to be. C'est la biotechnologie!

Tuesday, April 10, 2007

N-A-I-V-E, on a double word score

Monday's Globe had a short article on an MIT junior who plays competitive Scrabble and has also devised a Scrabble-playing program called Quackle. What was striking was the confident pronouncements of some Scrabble experts that a computer would never be able to beat them; of course similar pronouncements have been heard for checkers, chess, etc.

One expert opines:
Quackle has the right name for sure, because the whole idea of using a computer to play Scrabble is a quack of an idea

Another player makes the following analysis of a lookahead strategy
you're wasting your time because there's too much randomness ahead of you


IMHO, these domain experts are falling into the easy trap for domain experts: that what works for humans can't be computerized, and conversely what doesn't work for humans won't work for a computer. Humans are terrible at comprehensive lookahead strategies, but computers can do them flawlessly -- given enough time and compute cycles. This is sometimes a difficult idea to get across to domain experts -- both game players and biologists -- that it is more important to describe what your problem is than what they are certain are the aspects the computer can't do. Maybe some pieces can't be solved -- but maybe those pieces can be bypassed instead.

Of course, it does depend on your metric. The same critic of lookahead also offered
There's enough luck in the game that it's not really possible for pure word knowledge to defeat a slightly less pure word knowledge every time
. Now, for any game with a luck component it will be impossible for a program to win every time. I would agree with this -- and might even try (if I were doing this) to have the deeper levels of the lookahead use a much smaller dictionary (or more likely, some sort of word model along the lines of the word guessers in cell phones). The key isn't necessarily searching every possibility, but rather searching most of the most probable space.

On the other hand, who am I to talk. As an early exercise to learn Java I found a Reversi applet I could consistently beat & tried to improve its lookahead algorithm. Until I moved to more bioinformatic relevant problems, the program torture me -- because I could wallop my versions even harder than the one I started with! Any Hippocratic software oath of mine was badly violated!

Friday, April 06, 2007

woof WOOF!

I had offered to let the Omics! Omics ergonomics director scribe tonight's entry, since it covers a topic of which she has direct knowledge, but we ran into two problems. First, the keyboard is poorly engineered for the configuration of her digits. Second, she's too excited by the news to think straight. But, she did suggest the headline and sometimes that is the hardest part of all.

Dogs are amazing products of human selection. Size, color, shapes of various body parts, and even behaviors are distinctive to particular breeds. The size range of adult dogs exceeds that of any other mammal, as wonderfully illustrated by the
cover of Science
, as the issue announces the IGF1 locus as the major determinant of size in dogs.

This paper also illustrates how science can be a bit slow to get off the ground. About 15 years ago I heard a seminar from the senior author of this paper discussing the great promise of dog genetics to shed light on important medical and biological questions. In the meantime, there really haven't been much in the way of splashy dog genetics papers, though the community has been slowly building up a cache of tools. I think it is a reasonable expectation that this paper will be the head end of a series of papers unleashing the promise of the field.

After all, consider my diminutive assistant. . If you line her up with a Scottish terrier, while they are about the same size they have little else in common in their basic shape. Amanda has the flat face characteristic of her tong (The Ancient and Honourable Order of the Shih Tzu), whereas the Scottie has a distinctive snout which marks his clan. Her tail could honestly mistaken for a feather duster, whereas the terrier's has a less severe curl and isn't fluffy. A Scottie has erect ears; a Shih Tzu's are floppy. Each one of those characters could probably land a paper isolating them in a good journal. Perhaps even more striking would be the identification of a locus leading one of the sterotyped behaviors of a working dog. Half of Dr. Ostrander's seminar 1111 years ago (hey, this is an informatics blog!) was video of border collies herding sheep -- and ever since I can't watch a border collie playing with a frisbee or playing with kids and think of it except in terms of herding.

Many other domesticated animals have a large number of interesting breeds, but never quite as varied as dogs. Dogs were simply bred for more roles than cats, pigs, horses, cows or rabbits -- leading the selection of a wider variety of traits.

Yes, dog genomics promises to dig out some interesting biology, to fetch new insights into the genetics of morphology and behavior. When it comes to the secrets of the mammalian genome, we must not let sleeping dogs lie.

Tuesday, April 03, 2007

And then there were four

This weekend brought the news that in quick succession the last U.S. female veteran and last U.S. Navy veteran of World War I had died. Now only four veterans of World War I are known to reside in the U.S., three from the U.S. Army and one from the Canadian Army. Similar handfuls survive from the other participants. Despite its inevitability, it is striking when such a large cohort of people disappears. A huge chapter of human history loses all its living witnesses.

My own grandfather was a doughboy, though he never spoke about it. I have a few mementos of his service -- his helmet, a reproduction of the photo of his unit before it shipped out and a copy of his war diary lovingly transcribed by my mother from the disintegrating original. I am only one dereference away, but that is still a great distance.

World War I got scant attention in my grade school history classes -- a mention of the assassination which fermented it (which was probably the only mention ever of the Habsburgs!), a recitation of the events leading to U.S. entry, and perhaps an iconic picture of an American soldier in the Ardennes. The Armistice and the Paris peace talks would about round it out. Given the late entry of the U.S. and the still debated impact of its entry, and the fact that my history classes were universally parochial in their world view, it is not surprising.

On reflection, what is perhaps more striking is the minimal number of scientific or technological changes which are routinely traced to The Great War, particularly in contrast to the greater cataclysm it helped set up. While a number of innovations in human slaughter are routinely cited, about the only non-military influence I can immediately think of is the impact on aviation -- both by driving technical advances and by generating a large pool of pilots, who would later barnstorm across the U.S.

World War II on the other hand is permanently associated with a large number of technologies. Perhaps first on the list would be atomic energy, but radar and rocketry would be close behind. In the biomedical arena, WW2 caused a huge push forward in antibiotic production methods.

While the Genome Project seemed like big biology -- and it was, it was nothing in scale compared to the Manhattan Project or Apollo. Someday the last of these cohorts will leave this world. It may well pass unnoticed, as the prominence and certainty of military service are far beyond that of any civilian enterprise.

All of these individuals are of advanced age, considering that the war itself started 93 years ago this summer. What remarkable lives!

Sunday, April 01, 2007

Bits off the Wires

For almost as long as I can remember, I have been a compulsive reader. Our house had a daily paper and a number of periodicals, and I read many of them. Before my teens I was regularly reading two daily newspapers, as we subscribed to the one major Philadelphia daily and I delivered the other one. In college, I worked shelving periodicals, and so my breaks were often spent browsing magazines.

The Internet has of course expanded the reach of what I can access immensely. There are a number of regular sites I visit, but I also enjoy clicking through to things that look interesting. I use Gmail for my personal mail, and so my incoming mail triggers sidebar ads -- for example, if my mail mentions 'DNA', which isn't uncommon, there is at least one creationist site that routinely shows up. I also spend a lot of time Googling for useful information, which brings up all sorts of sidebar ads. Some are totally silly -- I can apparently buy DNA sequencers (and worn socks) on eBay, and others are varying degrees of odd. A few of the choicer items from recent searches are below.

An obscure English institute of learning convened a press conference in a small village to announce that a research team is close to identifying the gene responsible for the obscure genetic syndrome of 'squibism' -- sufferers apparently lack certain abilities this institution finds appealing. The item mentioned that it may be allelic to being a muggle, but I hadn't heard of that either. The spokesman mentioned that the institute is also investigating a number of unusual creatures capable of generating very high internal body temperatures, with possible applications in energy production or bioremediation. However, when pressed for details the spokesman vanished from the press conference.

A more frightening item concerns a precocious, but sinister, trio of children who are apparently playing around with genetically modifying a deadly fungus known as the Medusoid Mycelium. According to The Daily Punctillo, these Baudelaire siblings have apparently been associated with a series of unfortunate events, including arson and the loss of a research submarine. Luckily for all of us, a very dashing chap named Count Olaf has promised to track them down and separate them from their source of funds for these dastardly experiments. Given the facts presented, we should all have a Very Fervent Desire that this be accomplished as soon as possible.

Finally, I stumbled on a website for Samiam Biotechnology Inc., which is proposing to market foods based on chickens and pigs transgenic for green fluorescent protein. I always meant to try some modern biotech foods, but have never gotten the chance. However, this seems to be just a little too weird for me -- it's one thing to have food that is enhanced for nutrition or shelf life or flavor, but to modify its genes just to have fun eating under a black light? I think that's a bit much. No, I would not eat them on a boat, I would not eat them with a goat...