The Human Genome As a Cuisinart?
Last month's revelation that the human genome is comprised of about one-third the number of genes that scientists originally thought means one of two things: Either humans are creatures as simple as corn, or each gene is much more complicated than we thought.
Most researchers are going with the latter.
"If anyone found the basis for the pride of our species in the number of genes we had, they may have to rethink it," said Craig Venter, president of Celera (CRA), the company that published its map of the human genome in the Feb. 16 issue of Science.
But humans need not feel suddenly humbled because we have only twice as many genes as a fruit fly, and less than a pufferfish.
"It's OK that we're only (two) times a fly, because it's not the number of genes you have, it's the number of unique interactions they can have," said Jonathan Rothberg, president and CEO of CuraGen (CRGN) on Thursday at the Genome Tri-Conference in San Francisco.
"We have a feel-good understanding that it's OK to have only 35,000 genes," Rothberg said.
During the hubbub of press conferences and speeches surrounding the publication of the human genome maps, researchers declared over and over again that each gene is much more complicated than they once thought.
The genome is like a Cuisinart. It slices and dices in ways scientists are only beginning to understand.
But this was old news to an Israeli biotech company called Compugen (CGEN), said Lior Ma'ayan, vice president of commercial operations at Compugen.
"We have already invested ... close to one thousand man-years into this," he said. "Most people are still under the impression that this is a unique or remote problem. No, this is the rule, not the exception."
Researchers once assumed that one gene encoded one protein. They now believe that each gene encodes many proteins.
"It is already known that many human genes produce multiple proteins, each by using different parts of the gene to construct a protein," said John McPherson, co-director of the genome sequencing center at Washington University in St. Louis, Mo., after the publication of the human genome maps. "This is called differential splicing."
It's also called "splice variance," or alternative splicing and since 1996, Compugen scientists have been working to develop a proprietary mathematical model that predicts these pathways.
Located in between the genome and the proteome, this level is what Ma'ayan calls the "transcriptome."
"Compugen has in its hands the key to probably the most major post-genomic problem, which is understanding how genes translate into proteins," Ma'ayan said.
Simply looking at a gene is no longer enough. Researchers need to look at each pathway between gene and protein -- a path traveled by so-called messenger RNA.
By missing just one of these paths, a scientist could miss the key to a disease.
For example, in a paper published in the December 1999 issue of Nature Biotechnology, John Reed of the Burnham Institute in La Jolla, Calif., explains that a gene called "bcl-x" can splice one way to cause cell death, and another way to turn off programmed cell death, also known as apoptosis.
There is no serious way of looking into what happens at the protein level without understanding this intermediate level, and we understand this intermediate level," Ma'ayan said. "More than that we commercialize this level."
Some heavy hitters are already buying what Ma'ayan calls Compugen's "black box." Pfizer (PFE) of Viagra fame has been a customer since 1998, as well as Motorola (MOT) and Human Genome Sciences (HGSI).
Companies like Affymetrix (AFFX), which sell gene chips, Ma'ayan suggested, are playing catch up in the splice variance department.
On Wednesday, Affymetrix filed an 8K with the Securities and Exchange Commission saying it had found errors in the company's chip to read the genes of mice.
The errors come from the public database, called UniGene, where the company gets the mouse genome information, and have nothing to do with the manufacturing of their chip, said Anne Bowdidge, director of investor relations at Affymetrix.
But Ma'ayan of Compugen says his "black box" eliminates errors such as this one.
"They lack the ability to analyze this important phenomenon," he said. "There are so many things you need to model, including splice variance, and public databases don't go far enough because it's too complicated."
Affymetrix said in its 8K filing that the company's recent acquisition of Neomorphic allowed them to find the mistakes, and their expertise would allow them to catch such mistakes in the future.
"The acquisition has been able to prove its value," Bowdidge said.
Analysts don't seem worried about Affymetrix either. Mike King, an analyst at Robertson Stephens, reiterated his buy rating on Affymetrix stock, and issued a report saying he doesn't believe the glitch will affect the company's dominant position in the gene chip market.
If Affymetrix replaced the chips for every customer affected by the errors, they would have to spend $4 million, which King estimated would be little more than a drop in the bucket in the long run. Moreover, he said the situation is a halo for Affymetrix, highlighting it as a company that provides a warranty on its products.
But Ma'ayan said that the former Neomorphic scientists are just beginning to figure out what Compugen has spent the past four years perfecting.
"This is not something where you can just take three software engineers and one molecular biologist and get a solution," Ma'ayan said. "In due time they will see they are looking at problems, and I'm not sure how fast they can overcome them."