Analog Computers May Be Better at Body Simulations

Researchers from MIT design a new analog compiler (Yes, that's right: analog.) that could pave the way to efficient and accurate simulations of organs and biological organisms.

Kristopher Sturgis

Last week MIT researchers debuted their new compiler for analog computers at the Association for Computing Machinery's conference on Programming Language Design and Implementation. Their new compiler program was designed to take differential equations as input -- something that biologists frequently use to describe cell dynamics--and translate them into voltages and current flows across an analog chip.

In the realm of technology, analog systems have become a thing of the past. We've seen tape decks and TV antennas transition to mp3 players and streaming video, as the world around us becomes increasingly digital. When it comes to digital electronic circuits, events are often interpreted in a state of on or off, which is represented in 1s and 0s in binary arithmetic.

In the fields of synthetic and systems biology, the chemical reactions that lead to protein production in cells have often been compared to electronic circuits -- and for years researchers have been analyzing reactions in terms of these binary oppositions. However, in analog terms, a transistor can have an infinite number of states, which could represent an infinite range of mathematical values. In recent years, researchers have begun to rethink their approach to biological systems, positing that analog computers may be much more efficient than digital computers when simulating biological systems.

"Analog hardware devices are particularly well suited for performing biological simulations efficiently, but they're typically challenging to manually program," says Sara Achour, a graduate student in electrical engineering and computer science at MIT and first author on the paper detailing the work. "An analog compiler will allow researchers to automatically map their biological simulations onto analog hardware, without having to be hardware experts themselves."

The group tested their new compiler on five sets of differential equations that are commonly used in biological research. They found that on the simplest test set that only contained four equations, the compiler took less than a minute to produce an analog implementation. The most complicated set, consisting of 75 differential equations, was completed in less than an hour--something that could never have been accomplished manually by hand.

"At the moment, we are focusing on automating what is mostly a manual process," Achour says. "There is a lot of potential for improving the performance of the compiler that we are excited to pursue in future work."

Once the compiler has a promising algebraic redescription of a set of differential equations, it can begin mapping elements of the equations onto circuit elements. As they move forward with their research, Achour says the group is committed to improving the hardware to help produce highly accurate and efficient analog simulations for biological systems.

"Since our initial system already compiles dynamical systems that have been published in biology journals, we are focused on improving the fidelity of the hardware model our compiler uses," she says. "We are working on adding support for dynamic range mapping, which involves mapping value ranges in the dynamical system to sub-ranges of the current or voltage, and reasoning about the effects of this mapping in an automated manner."

The hope is that this group will continue to explore the untapped potentials of analog components, to better understand how they can effectively drive biological systems forward.

Kristopher Sturgis is a contributor to Qmed.

Like what you're reading? Subscribe to our daily e-newsletter.