Perfect Storm

There is some interesting lack of precision, to the point of “error” on the way some concepts are dealt with by human language.

Take, for instance, the concept of drug half-life, i.e. the time that it takes for the concentration of a drug to drop to half (drug concentrations in the blood are normally modeled through exponential decay), it is conceived as a property of the drug – people talk about drug D has an half-life of H hours – but it is really a property of both drugs and individuals (actually is much more complicated than that, we could repeat the argument).

And no, this has not only to do with statistical deviations that are acceptably approached by the drug only.

As example, there is a study about the pharmacokinetic properties of Sulfadoxine-Pyrimethamine (a widely used cheap antimalarial). In this study, there is a big deviation for half-life (and other parameters) for the children between 2 and 5 years. The study concludes that “dose recommendations need revision” for that group. To put in another way, half-life (and other parameters) is not (only) a function of the drug.

Now, I am not suggesting that the concept of half-life tied just to the drug should be thrown away. I am just speculating why it is framed as a function of the drug only, as clearly that is not the case.

First there is probably historical inertia: The concept was first framed that way at a time that it seemed that half-life was only dependent on the drug and it stuck by “memetic” inertia.

But, much more importantly, it is still there because, it is both less expensive (it is easy to express half-life as a function of just the drug, than other parameters which might be still crucial in some situations) and still meaningful enough in many contexts (for instance, expressed as a function of drug it is still useful to compare the half-life of Artemether – short – against Sulfadoxine – long – for many kinds of reasonings). Even when the most economical concept entails some errors it might still be practical. The problem only arises when its simplicity has bad consequences (in this case, having wrong drug doses)… but, in certain contexts, it might be a problem, a serious problem (See my previous text about the notions of resistance, tolerance and sensitiveness for an example).

It all depends of the discourse context, but one should be careful.

As an anecdotal example if you are seriously ill and a doctor prescribes you a pill, do you prefer to hear “this will cure you” or “this will drop the parasite load at a rate of 1 order of magnitude per hour starting 3 (90% CI of 2.5 – 3.5) hours after intake. Parasite load is expected to drop to 0 in 10 hours”?

The problem arises when the cognitive bias of the simplicity of “this will cure you” gets into more rigorous contexts.

This has implications on the computational modeling of concepts. The tradition in computer science it to “dig down” to the “real meaning” of concepts. In that sense simpler explanations are deemed “wrong” (and should be rewritten in terms of “correct” conceptualizations). Maybe a different strategy is needed, one that takes some linguistic and cognitive economy to computational systems (while still maintaining rigorous and precise reasoning and conceptualization when that is needed – like human languages can do).

I am going to stop here, but I think that one of the problems that impairs mathematical modeling is the application of the “certainty of numbers and formulas” to non-rigorous concepts. Then you have the worst of both worlds: an authoritarian argument (mathematics is a foundation for authority. “The numbers prove it”) based on modeling vague, imprecise and wrong concepts. But that is a topic for a another post.

Being a computer scientist with a strong interest in languages (languages in the broadest sense possible: programming, natural and cognition related issues), I am in an holy grail quest for a programming language that:

First and foremost allows me to express my computations in a way that is close to the problem domain (as opposed to close to the machine). As I am working in a biology setting that means being able to talk about concepts around genes, epidemics and pharmacology in my programs. I don’t want to think about CPUs, memories and things like that when I am coding. Prolog and Lisp are good examples here. I also need programs that can evolve over time as knowledge changes, I need strong metaprogramming and Domain Specific Language facilities.

Unfortunately I have a couple more requirements coming from the day to day reality…

Real world: I want a language that interacts with existing libraries and that I can easily make available to other people to use, inspect and change. I need Bio* libraries, graphics plotting libraries. I my personal case I decided that I want to work inside the JVM, so I need a language that works in the Java world (Jython, JRuby, Scala, Groovy, … Java).

Software engineering: Programs have to be easy to maintain and debug. I guess there is no way around explicit typing on the debug and tool construction front.

Ridiculous religious fanatic quest? Yes, it might be, but I am pursing it.

The truth is that we are not far away from this grail.

Scala is almost there. Lacks metaprogramming and things like type inference are a bit amateurish (compare it with CAML).

JRuby is maybe there, I could live with it, I guess. The lack of explicit typing will make things difficult in the long run on the software engineering front.

I decided to give a final try to yet another language: Groovy, and up to now it is going very OK. Seems to nail all the fundamental points. I especially love the effort on good metaprogramming facilities.

I decided, for pragmatic reasons, that after this one I will stop my pursuit for the grail. If Groovy proves a blunder of some sorts I will revert to JRuby and carry on.

I am currently developing the Biopython module for population genetics and genomics (by the way, you are invited both to help with the development and to make suggestions – maybe based on your needs – for new features).

On the current (1.44) version of Biopython, a GenoPop parser and code to deal with FDist (a Fst outlier method for selection detection) is available.

It is my pleasure to announce that coalescent simulation (in the form of support for the SimCoal2 simulator) is currently available on CVS and will probably be out on the next public version. This includes, code, test code and DOCUMENTATION. This means you can now do coalescent simulations from inside Biopython (many demographies and markers supported).

Future plans for Bio.PopGen include statistics (the meat of the module, actually) and HapMap support, among others.

Need any feature? Just ask. I cannot promise it, but I will try to address user requests in as much as possible

I really did not have an holiday, but I stopped posting for a while.

But I want to talk about another “holiday”: Scala.

I have spent a couple of months with Scala: A functional-OO programming language done from the scratch with the JVM in mind, with a nice, smart community.

I actually decided to stop my efforts on Scala and decided to go back to explore the Ruby way… The reasons:

1. No metaprogramming facilities. This comes from ML, I suppose. But Ruby has it and many “old school” elegant languages have it (Lisp, Prolog). It is possible to be elegant (in fact I would contend that in many settings it is a requirement) with metaprogramming.
2. There seems to be some difference in the semantics between compiled and interpreted. I only compiled, but the interpreter could add new variables to its local scope (as it really needs it) but the compiler couldn’t. While one might argue that that is excessive flexibility coming from the scripting languages camp, but I actually had to, on a compiled program, to create new classes which would include traits that would be dependent of need of the user, and this cannot be done. If one has many traits, it has to compile a priori all the trait mixins desired, they cannot be defined at run-time in a compiled environment (contrast this to JRuby or even JPython). This is actually metaprogramming lacking part 2.
3. Type inference: Scala type inference might seem clever, but, compared to CAML it is not. Sometimes the compiler is not able to infer the types and the user has to explicitly declare them. CAML was always capable (at least in my cases) of complete type inference.
4. Information sources are scarce. The mailing list is reasonable, but sometimes questions get unanswered and there is no other source (other than inspecting the source code). This will sort out if there are more people using it – and more books like the Artima ebook.

Decent metaprogramming in a runtime setting would be my main requirement, but in the current Scala status, one can only have it though the typical Java way: execute the compiler, link a jar, not elegant…

Regarding Ruby, I would like to have some form of strong and explicit (or inferred) typing. I would imagine that the requirements of metaprogramming flexibility and typing are contradictory, but, at least, some kind of optional (but standard) annotation for input/return parameters would allow avoiding some debugging nightmares of not having the compiling helping with types and would also allow for smart code editors to do all that fancy completion that is possible with explicit typing.

[This was initially posted - with modifications and additions - on Artima as a comment]

One of the most delightful things in bioinformatics is the possibility of working with people with really different mindsets. Surely CS geeks are amazing, and everyday I feel that my original background is really a comparative advantage, but, from where I look, nothing beats being in an environment with scientific and cultural diversity. But, lets talk some geekiness now:

A couple of years ago, I did a population genetics simulator in Caml. It was really flexible, allowing for many demographic and genomic scenarios, mating rules, selection… really flexible. I never got to try to publish it because there are many good simulators around (I suggest simuPOP, if you are looking for one) and it would take some time to make it robust and documented for public exposure. But, the interesting part is, when I went to my MSc supervisor (an “old-type” biologist) and after a very exuberant explanation on how flexible the simulator was, he added only one comment: That is all very well and good, but you did not show me the easy to use graphical interface!

Fast forward a couple of years… With regards to a DSL to model drug resistance in the context of infectious diseases that I am developing, I went to my PhD advisor (a population geneticist, malarialogist, biostatistician who knows how to program in C), showed him my rough prototype and he said: People will be able to read this, but, to interact they will want an easy to use graphical user interface. To be honest, this time, I was expecting the comment (I am living in the middle of experimentalists long enough to have learned something). I have no expectations, for my DSL, that domain specialists will write it (well, maybe a couple of them will, if things pick up). If I end up giving my system away to domain specialists, it will have to have a easy to use interface, there is no escaping from that.

Well, DSLs (at least in Scala and in Ruby) have an underlying OO model. Which, most of the times is neither complex nor big. I am starting to suspect that it won’t be too difficult to automatically generate an easy to use interface to input in a “nice” way what could be rendered as DSL programs (or object instances and relationships, if you prefer to look at it that way). For embedded DSLs, which have the whole expressive power of the host language available, that would be unfeasible to do completely. But, at least part of it could be automated. Obviously this idea is not new at all, this is just a rehash of what Lift or Rails do for databases.

I am aware that graphical programming languages never went too far (I actually dislike them), but the scope and context here are completely different, different premises apply. This might be one way of lowering the barrier to rigorous modeling to a wider crowd.

I would like to make a preemptive defensive comment on the new population genetics module.

I am, for now, the sole author of the code that is there (although, in future versions there will be at least code from another person. By the way, if YOU want to participate, your 100% welcome). Although the code is mine there was a lot of help from Peter Cock, one of Biopython’s core developers. Without him, this initial groundwork would not have been possible.

Now for the preemptive defense :

If you look at the module, it has very little functionality included. This is a very deliberate strategy to start small and grow slowly. I am expecting for some feedback (which will be very little, I am sure). I want to grow in small steps, including as much feedback as possible. Test code and documentation have to exist before releasing anything to the public.

In the pipeline there is code for coalescent simulation, statistics (including code supplied by Ralph Haygood, that I am joining with my own) and HapMap. If you are interested in early access to any of this code, please give me a shout as most of it already exists. Alpha testers are more than welcome .

I am currently trying to model antimalarial drug behavior in order to understand the spread of drug resistant malaria. Generally speaking, malaria strains are more or less tolerant to a drug depending on the quantity of drug that is necessary to kill an infection. In theory, a totally resistant infection will survive any treatment, a totally susceptible one will only require small levels of drug to be cleaned.

I see the word drug used in two different ways (for the readers of this blog that are specialists, in some form, on issues regarding drugs, particularly pharmacokinetics, if you see any thing particularly wrong, please do inform me): For instance, SP (Fansidar) is a drug, composed of two drugs (Sulfadoxine and Pyrimethamine). I will use drug for SP and compound for S and P (as active compound seems to be used).

Antimalarial drugs work mainly in the blood stream against asexual parasite forms.

In the blood, compounds have a certain concentration. With time, the body gets rid of compounds (thus the concentration of a compound goes down with time). The concentration of compounds is normally (but not always) modeled using an exponential decay function, being the fundamental parameter the half-life, i.e, the time that it takes for the concentration of a compound to drop to half.

Two other important concepts for drugs that are not taken intravenously (like cheap antimalarials which are oral), are

1. Bioavailability, i.e. the fraction of the compound that actually reaches the circulation. It seems that one of the problems with counterfeit drugs is low bioavailability. Bioavailability is normally discussed in terms of AUC (Area Under the Curve. Being the curve related to the plot of drug concentration against time). I will model it in terms of maximum concentration, half-life and the time it takes to reach maximum concentration in the blood, which by the way is the next concept…
2. The time it takes to reach maximum concentration in the blood, i.e. the time from ingestion to circulation in the blood at maximum concentration. I suppose this time frame has a technical name, but I don’t know it (if you know, drop me an email our comment, please).

Now, back to computational modeling:

A big objective is declarative programming. Preferably a program that can be read by domain specialists (biologists, MDs, biostatisticians, …), with that in mind…

Currently, a computer program in Scala to model drugs look like this.

Compound create "Sulfadoxine"
Compound abbreviation "S"
Compound half_life 116 //hours
Compound bio_availability 408 //1mg to nanoM
val Sulfadoxine = Compound prepare
 
Compound create "Pyrimethamine"
Compound abbreviation "P"
Compound half_life 83 //hours
Compound bio_availability 34 //1mg to nanoM
val Pyrimethamine = Compound prepare
 
Drug create "SP"
Drug includes Sulfadoxine quantity 500
Drug includes Pyrimethamine quantity 25
val SP = Drug prepare

Discussion:

• I am using the “object companion” pattern a lot. The idea is that all “stateful” mess is stored “prepared” in the object (which is the DSL source). When the prepare method is invoked in the object a class (with only immutable vals, very lovely for those of you who are functional programming enthusiasts) is created.
• Notice the dependence on operator precedence on Drug includes quantity (there is not really one, strictly speaking, but assume there is). I would really like to have, per class the ability to define operator precedence, other than not based on dictionary order (à la Prolog).
• I don’t like the val SP = Drug prepare. It is too verbose and too geeky. I would prefer just Drug prepare. I believe that this is possible in Scala as at least at the interpreter level (as the Scala interpreter does it), but I still don’t know how. The idea would be that a val named SP would be added to the local scope in some way. For those computer inclined readers that think that I am being too pedantic and nit-picking, I just have one thing too say: I am really trying to make the system the most pleasant possible to non programmer types, and I think my proposal does not sacrifice elegance and generality (although I would recognize the non-explicit name creation is “strange” – but, hey, the Scala interpreter already does it!)

Caveat emptor, big one: Although drugs (compounds) are discussed in terms of half-lives, bioavailability, etc… these properties are actually not of the drug but of the interaction between the drug and the individual. Making them drug properties only is a “cognitive abuse”, although it has its uses. For instance, my advisor, after looking at the language, was talking about bioavailability for counterfeit drugs, for children between 2 and 5 years. A great example that they are not properties only of the drugs but also, at least, of individuals (and not only that, for instance many drugs are more bioavailable if there are taken in conjunction with, say, fatty foods).

A proper, precise, computational modeling of drugs would be a gigantic undertaking. I have a different approach: Modeling as close as possible to the average domain discourse and hook, in some way, the necessary precision, should the need arise. It is worth noting that “incorrect”, “imprecise” modeling is enough for many tasks.

My PhD thesis is around the spread of drug resistant malaria. The idea, from my supervisor, is that between sensitiveness (i.e. a parasite strain that is killed by a drug) and resistance (i.e. a parasite strain that that survives treatment) there are intermediate forms, tolerant forms, that are able to resist sub-terapheutic levels of treatment. The more I read, the more I like the idea. In fact, I think there is no thing like resistance or sensitiveness, there are only various degrees of tolerance. Let me explain:

Imagine a parasite (a single one, one in an infection of 10000000000 parasites), this guy is resistant to chloroquine and is exposed, by chance, to a MASSIVE amount of chloroquine. He will die (“he”, by the way, is a “it”, as we are talking about asexual forms. Although P. falciparum also has sexual forms). This interpretation is supported for instance, in the fact that chloroquine “resistant” infections can still be treated with chloroquine as long as the host has accumulated immunity (as in adults in malaria endemic areas). There is an “add-up” effect of the drug plus the immunity. Put it another way, a “resistant” strain is still partially affected by the drug.

The converse is also true, a sensitive (I should be saying highly intolerant) parasite might survive chloroquine treatment if either the drug level is sub-therapeutic or, shear luck of not crossing in the way of the drug.

There are good reasons for the existence of resistant and sensitive words though, I can think of at least two: First, from the outcome perspective – either the infection is cleared or not. Second from a cognitive and linguistic perspective: it is easy to talk and think about “black and white” sensitive and resistant than “gray toned” level of tolerance.

But I think that “resistant” and “sensitive” create a cognitive bias that undermines the ability to understand the underlying biological processes.

I prefer the notion of “tolerance” that, for me, is tightly associated to the notion of drop in parasite load per unit of time. A resistant infection, exposed to a drug, still has some drop in parasite load, but that is not enough to offset multiplication of surviving parasites (at least to bring the parasite count to 0). A tolerant infection, exposed to the same quantity of drug, has a big drop in parasite load, enough to offset the multiplication of surviving parasites that most probably will be eliminated soon.

The idea of tolerance and drop in parasite load also goes well with the importance of drug concentration.

It turns out that my previous Scala idea to model environments for arithmetic expressions doesn’t fly. Solution? Go to the Scala mailing list and ask for opinions, there is always a lot of helpful, highly skilled people willing to help. What follows is my account of things based on the help that I have got.

My initial solution was to create a type like this:

type Env = String => Double

Doesn’t work as types have to be defined inside objects or classes. This will, by the way, work on the interpreter as the interpreter is really running inside an anonymous class (so type definitions will be attached to that anonymous class).

I then tried 2 traits as alternatives:

trait Env {
 def apply(name : String) : Double
}

and

trait Env extends Fuction1[String, Double]

When I used any of those like in:

val env : Env = (x : String) => x match {case "S" => 1.0}

I had a typing error complaining that String => Double is different from Env. But is it really? Aren’t the two traits above the same as String => Double? Well, they are, but Scala uses nominative typing, not structural typing. So after you give that type a name it will be different from equal structural types that have different names.

I ended up with with a type inside a companion object. The companion object pattern seems to be widely used in Scala. I still don’t have a opinion if that is generally good or bad.

In any case, I don’t like the idea that types cannot be defined top level. I don’t see any advantage in disallowing this.

I will document my effort in learning Scala. The main objective is to help in building up as much content as possible on the Web about Scala. Caveat though: I am learning, therefore what I document here might be rubbish.

The problem: I need an arithmetic expression parser to be able to deal with drug isobolograms like these 2 taken from Wang et al.:

Very roughly speaking, the concentration “curve” that you see is when the drug combination becomes active against the parasite. In this case the compounds are Sulfadoxine (SDX) and Pyrimethamine (PYR) used to combat P. falciparum malaria.

I want to be able to build expressions for those curves. These expressions run in contexts, that is, if one “curve” (here approximated by a line) is 133*SDX + 2000 – PYR then there is a need for having a variable SDX and another PYR.

[People with a functional programming background might immediately recognize the typical newbie exercise of doing an expression evaluator... In fact you can find it in various Scala documentation... and Caml]

So I need, what is called an environment, a store of mappings from symbols to values. Something like

{case "PYR" => 1800.0 case "SDX" => 5.0}.

First problem: Doubles and Ints. There seems to be no way to specify that the return is a Number. I would like, sometime, to use Numbers irrespective of the specific subclass. I would like to do something like:

type Exp : String => Number

The problem is that one cannot have type at all, unless inside a class. But, but, but the type is not visible outside the class as a type (i.e. just for the type info, without the need to instantiate to access). Maybe putting it into an object? Did not try it, but it would be clumsy.

I ended up with this solution for now:

trait Env {
  def apply(name : String) : Double
}

Can live with it. The Double instead of Number hurts more.

Again, I am beginning. Please don’t be too harsh on me

By the way: It would be nice to have Scala support on GeSHi so that WordPress blogs could render Scala code beautifully. No, I am suggesting only, I don’t have the time to actually do it myself.