Duke Chemist Describes "Fluorescent" Method For Computing With DNA

The modification is the Arrayed Primer Extension (APEX) method, which allows DNA molecules to undergo reactions more precisely by binding them to glass. The Duke team has demonstrated how to also use APEX to perform computer-like digital calculations. A special chemical feature of APEX then displays solutions to those calculations as tiny but visible colored fluorescent spots.

"The human eye is sometimes a better detector than a lot of fancy instruments," said Michael Pirrung, a Duke professor of chemistry, in an interview. "You can look at this and say: 'This is a digital signal.'"

In a talk during the American Chemical Society's late-August national meeting in New Orleans, Pirrung described using the APEX method to compute certain kinds of problems known as "satisfaction" (SAT) that require solving equations with enough variables to challenge the most powerful supercomputers. Computing with DNA can best the efforts of supercomputers on such problems because a huge array of molecules can be unleashed on the calculations simultaneously in a small amount of space.

Pirrung is among the pioneers of automated techniques for conducting tens of thousands of chemical experiments at once on proteins, DNA or other molecules bonded to dime-sized silicon chips.

He said his research in DNA computing grew out of his work with the APEX method, which he developed with John Shumaker, Tom Caskey and other collaborators during a sabbatical at the Baylor College of Medicine in Houston, and also at the urging of his former post-doctoral student Richard Connors, who now works for Pharmacopeia, a pharmaceutical firm in Princeton, N.J.

DNA's use as a calculator stems from the fact that each of these massive molecules is made up of four different kinds of nucleotide bases - often abbreviated as A,C,T, or G.

These bases are arranged in two different strands that are supposed to stick to each other precisely, because each A in one strand is attracted to a T in the other strand, and likewise each G to a C.

Because A pairs up with T, and G with C, those two separate sets of bases are considered "complements" of each other. If A instead pairs to a different base than T, and likewise G to a base other than C, those pairings are considered mismatches.

The APEX method was actually developed for other uses than DNA computing. One example is genetic sequence analysis. It makes use of a common laboratory procedure in which natural cellular enzymes called DNA polymerases help synthesize additional DNA bases at the end of one of the molecule's two existing strands.

Working to the right of this to-be-extended "seed" strand, which is known as the "primer," the enzyme is supposed to manufacture additional bases that are complementary to prefabricated bases on the other strand, called the "template."

But mismatches can occur during the binding step between primer and template. And the APEX process is designed to eliminate these imperfections by premanufacturing extra starter lengths of both primer and template base sequences that are all the same and complementary with each other.

Large numbers of these identical primer-template sequences are also bonded onto a glass slide by their primer strands. That's because DNA manipulations are more efficiently done on a surface than in a liquid solution, and also more easily automated.

While each of these primer-template lineups is chemically identical, APEX designers next introduce some variety by adding four additional template base combinations that differ from one strand to another.

In response, the DNA polymerase can extend each corresponding primer strand by synthesizing four new bases, but only if those additional bases will be complementary matches to their guiding variable template bases. If any mismatch starts to form in the primer sequence, the enzyme will halt its work.Watching researchers will know whether the matchup is perfectly correct because DNA bases used in APEX reactions are also tagged with a molecule called fluorescein, which emits a colored fluorescent signal only if the enzyme has completed its sequencing.

While the APEX method's purpose is to improve the accuracy of DNA matchups, Conners saw it could also be used as a "high fidelity" DNA computer to solve a class of SAT problems called "NP-complete."

To tackle NP-complete problems, however, Conners and other members of the Duke team - including Pirrung and post-doctoral research associate Amy Odenbaugh - had to first further improve the accuracy of the APEX method with some additional chemical tinkering.

The APEX process can be used as a DNA computer because glass-bonded bases in the variable region of each primer chain become the "hardware" for computational chemical reactions. Whether or not this hardware emits a signal - the fluorescent spot - depends on whether its "software," the corresponding variable region template bases, is completely complementary with the primer.

Each colored spot tells scientists that there was no error in the reaction there. Since they already know the makeup of bases in the variable part of the template, the signal also assures them that they know the identities of the corresponding variable primer bases.

Many thousands of template bases, all suspended in solution, thus become the raw ingredients for solving an SAT problem. The array of primer bases, all arranged on different parts of a glass chip whose locations are known to scientists, are then dipped in that solution.

Wherever they find signals, they will know they have part of one of the many solutions to an SAT problem. It will be a partial solution because one APEX reaction must be carried out for each clause of an SAT equation.

In the binary language used by conventional digital computers, a non-signal would be recorded as a 0 and a signal as a 1, the only two digits that conventional computers calculate with.

Pirrung said the Duke-modified version of the APEX method thus becomes an improved way to do the kind of DNA computations described in a seminal article by Leonard Adleman in 1994, and refined for SAT calculations by Richard Lipton in 1995.

The problem with Adleman's method is that it involves conventional DNA copying reactions done in a series of conventional test tubes, Pirrung said. "There are all kinds of manipulation problems that you really ought to worry about," he added. "One little twitch of a finger and there go half your data points, because you've lost your DNA. And the DNA can also stick to glass in an uncontrolled way.

"Whereas, if you've attached the DNA to glass in a specific way, in the spatially directed way that we do in our experiments, then you know where it is. So it ought to be very straightforward."

Other researchers from nine different universities are also studying the use of biomolecular computing to solve such problems in a large, separate National Science Foundation-funded research effort whose principal investigator is John Reif, a Duke professor of computer science.