3.4 Programming Strategies

In order to give you, the beginning programmer, an idea of how programming is done, let's see how an experienced programmer goes about solving problems by giving a couple of instructive case studies.

Imagine that you want to count all the regulatory elements^[1] in a large chunk of DNA that you just got from the sequencing lab. You're a professional bioinformatics programmer. What do you do? There are two possible solutions: find a program or write one yourself.

^[1] A regulatory element is a stretch of DNA used by the cell in the control of a coding region, helping to determine if and when it's used to create a protein.

It's likely there is already a perfectly good, working, and maybe even free program that does exactly what you need. Very often, you can find exactly what you need on the Web and avoid the cost and expense of reinventing the wheel. This is programming at its best—minimal work for maximal effect. It's the classic case of the experimentalist's adage: a day in the library can save you six months in the lab.

An important part of the art of programming is to keep aware of collections of programs that are available. Then you can simply use the code if it does exactly what you need, or you can take an existing program and alter it to suit your own needs. Of course, copyright laws must be observed, but much is available at no cost, especially to educational and nonprofit organizations. Most Perl module code has a copyright, but you are allowed to use it and modify it given certain restrictions. Details are available at the Perl web site and with the particular modules.

How do you find this wonderful, free, and already existing program? The Perl community has an organized collection of such programming code at the Comprehensive Perl Archive Network (CPAN) web site, http://www.CPAN.org. Try exploring: you'll find it's organized by topic, so it's possible to quickly find, for example, web, statistics, or graphics programs. In our case, you will find the Bioperl module, which includes several useful bioinformatics functions. A module is a collection of Perl code that can be easily loaded and used by your Perl programs.

The most useful kinds of code are convenient libraries or modules that package a suite of functions. These packages offer a great deal of flexibility in creating new programs. Although you still have to program, the job may be only a small fraction of the work of writing the whole program from scratch. For instance, to continue our example of looking for regulatory elements, your search may turn up a convenient module that lists the regulatory elements plus code that takes a list of elements and searches for them in a DNA library. Then all you have to do is combine the existing code, provide the DNA library, and with a little bit of programming, you're done.

There are lots of other places to look for already existing code. You can search the Internet with your favorite search engines. You can browse collections of links for bioinformatics, looking for programs. You can also search the other sources we've already covered, such as newsgroups, relevant experts, etc.

If you haven't hit paydirt yet, and you know that the program will take a significant amount of time to write yourself, you may want to search the literature in the library, and perhaps enlist the aid of a librarian. You can search Medline for articles about regulatory elements, since often an article will advertise code (an actual program in a language like Perl) that the authors will forward. You can consult conference proceedings, books, and journals. Conferences and trade shows are also great places to look around, meet people, and ask questions.

In many cases you succeed, and despite the effort involved, you saved yourself and your laboratory days, weeks, or months of effort.

However, one big warning about modifying existing code: depending on how much alteration is required, it can sometimes be more difficult to modify existing code than to write a whole program from scratch. Why? Well, depending on who wrote the program, it may be difficult just to see what the different parts of the code do. You can't make modifications if you can't understand what methods the program uses in the first place. (We'll talk more about writing readable code, and the importance of comments in code, later.) This factor alone accounts for a large part of the expense of programming; many programs can't be easily read, or understood, so they can't be maintained. Also, testing the program may be difficult for various reasons, and it may take a lot of time and effort to assure yourself that your modifications are working correctly.

Okay, let's say that you spent three days looking for an existing program, and there really wasn't anything available. (Well, there was one program, but it cost $30,000 which is way outside your budget, and your local programming expert was too busy to write one for you.) So you absolutely have to write the program yourself.

How do you start from scratch and come up with a program that counts the regulatory elements in some DNA? Read on.