File Compression: New Tool for Life Detection?
|In 1996, John Grotzinger of MIT published a paper in the journal Nature showing that stromatolite-like structures could be formed through a simple chemical process, without the help of microorganisms.
To those interested in early life on Earth, stromatolites are among the most intriguing life forms known. Billions of years ago, they may have been the dominant form of life on the planet. Some scientists speculate that, perhaps, similar organisms once populated Mars.
Strictly speaking, stromatolites are not organisms, but rather layered structures built by colonies of microorganisms, the first great wonder of the ancient world. Until a few years ago, whenever geobiologists found an ancient rock that looked like a fossilized stromatolite, they figured it was a stromatolite. But in 1996, John Grotzinger of MIT published a paper in the journal Nature showing that stromatolite-like structures could be formed through a simple chemical process, without the help of microorganisms.
So how does one separate the wheat from the chaff, the true stromatolites from the fakes?
One method is to examine the suspect rock with a microscope, looking for visual evidence of microorganisms. But as researchers who study ancient terrestrial rocks- and one notorious Martian meteorite - have discovered, it isn't all that easy to tell, just by looking at shapes, whether or not a microscopic blob in a rock was once alive.
Stromatolites are handy in this regard because their characteristic layers, although formed by microscopic organisms, are visible to the naked eye. So if there were an easy way to distinguish biological stromatolites from non-biological layered structures that merely look like stromatolites, it not only would help in understanding the evolution of early life on Earth, it might also prove a useful tool for detecting evidence of ancient Martian life because stromatolites could be detected easily by a rover's camera.
|Frank Corsetti (shown above) proposed compressing digital image files of ancient rocks to look for signs of early Martian life.
Frank Corsetti, who is with the Department of Earth Sciences at the University of Southern California (USC), and Michael Storrie-Lombardi of JPL, think they may have found a way. Their approach is simplicity itself: Create a digital image of the rock; then compress the image file. The more the file shrinks, the more likely it is that life was responsible for building the layers.
Corsetti and Storrie-Lombardi began by collecting a series of TIFF images, some believed to be of biological stromatolites, others believed to be of non-biological layered structures that merely look like stromatolites. When they compressed the images on a Macintosh computer using the standard UNIX compression utility gzip, they found that the images of biological stromatolites were more compressible.
An RGB image file, like those used by Corsetti in his analysis, is comprised of a series of data bytes. Three bytes - one red, one green and one blue - represent each pixel in the image. The numerical values of the bytes indicate the color of the pixel.
When gzip compresses a file, it looks for multi-byte patterns in the data. It assigns a number to each pattern it finds and maintains a table that pairs patterns with their corresponding numbers. Each time a pattern recurs in the original file, instead of storing the pattern, the compressed file stores the number that represents the pattern. In this way, a pattern tens or even hundreds of bytes long can be represented by a number that is one, two or three bytes long. This is why the compressed file is smaller.
Corsetti explains that the biogenic images are more compressible because they're more predictable or redundant, which can be considered a form of biologic complexity.
This notion may seem counter-intuitive. Initially, Corsetti says, "I expected that the biotic one would be harder to compress, and that the abiotic one would be more regular, easier to compress. So we did the analysis - and it came out reversed."
Resolving this apparent contradiction requires a realignment of one's ideas about complexity. In this context, "complex" doesn't mean "complicated"; rather, it means, "patterned," less random.
|"The thing that maybe has turned a lot of scientists off to the idea of complexity, is that there has not been any universal or simple way to quantify what you mean by complexity." -Robert Hazen
Credit: George Mason University
"A computer file that's complex," says Corsetti, "would have a lot of non-redundant, or random, features in the data. Patterns would be hard to see in something that is more random. Something that's very redundant would have a lot of repeated patterns, and would be more compressible, and therefore less complex."
Although biological stromatolites and non-biological stromatolite-look-alike structures appear similar to the human eye, the biological origin of stromatolites makes them more ordered, more highly patterned. And it is this patterning that, while hard for the human eye to discern, is readily detected by the compression algorithm. Non-biological stromatolite-like structures are more random, less patterned and therefore less compressible.
Robert Hazen, who is with the Carnegie Institution of Washington, has spent the last several years studying complex systems, both living and non-living. He's intrigued by Corsetti's findings.
"The thing that maybe has turned a lot of scientists off to the idea of complexity, is that there has not been any universal or simple way to quantify what you mean by complexity. The sort of thing - I know it when I see it - isn't very useful," he says. His interest is piqued, however, "any time someone comes up with a way to look at a system and says, let's see if we can't in a purely mathematical impartial way see if there's a difference in the quantifiable structure" of a living versus a non-living system.
"It's an amazingly simple idea," says Hazen of Corsetti's work. "It's one that does not pre-judge anything genetic about the structure. It just says, can we differentiate between those stromatiform objects, which are not biological, and the stromatolites, which clearly are? And, lo and behold he does this and finds that - wow! - there's a difference."
Corsetti is the first to raise a cautionary flag about this work. "This is very preliminary. I'm not saying we have the answer. In fact, I don't believe we have the answer. What we have are some intriguing preliminary results."
So far, he and his colleagues have examined only about 20 images. They appear to cluster into two groups, one biotic, the other abiotic. "We need to do a lot more work to see, is this a spectrum, an interaction between abiotic and biotic things, or is it truly a signature [that can distinguish between the two]? And right now I'm not ready to say one way or the other, if it's a signature or not. What I am willing to say is that it deserves further research."
That research is already underway. Corsetti and Storrie-Lombardi plan to examine many more images in the coming months in an effort to determine just how useful the technique may prove.
"Our goal is to eventually have some kind of data base or website, where we could provide the protocol and other people would analyze their stromatolites in the same way and see what they come up with.
"It would be really neat if people could send us an image - they will determine to their eye, based on the microscopic structure, whether they think it's biotic or abiotic, and send it to us as an unknown. We could do the analysis, see where it falls, and then see if it matches with what they found. And that's kind of cool because it adds a certain double blind aspect to the study, which at this point we don't have."
When asked if he planned to apply his technique to close-up images of Mars rocks sent back by the upcoming MER missions, Corsetti said, "Oh sure, I'd love to look at those! But we're not ready to say, hey, we can make a comment one way or the other."
Try an image comparison here to test its approximate biogenic complexity
Related Web Pages
Early Life on Earth: Stromatolites
How File Compression Works