When Francis returns from his trips to Bangkok, it's to Brisbane, a community on the San Francisco peninsula, midway between the city and its airport. His home and workplace both looks eastward across the bay: towards Oakland and, beyond that, America. His home is on a hill and lined with Chinese paintings. His office is by the shore, in black glass.
He huddles weekly with his senior colleagues: VaxGen's vice-president, Dr Phillip Berman, and its chairman, Dr Robert Nowinski. Berman, aged 49, is a molecular biologist. He's heavy set with curly hair and has laboured on the science for 15 years. Nowinski, 52, is bald and wears glasses. He's a biotechnology entrepreneur from Seattle. His main claim to fame is having founded and sold a company, ICOS, which boasts Microsoft's Bill Gates as an owner.
The key document at many of their sessions is a "special issue" of a prestigious journal, called Aids Research and Human Retroviruses. It's dated last October. Twenty papers are inside and they're a rave for VaxGen's ideas. Dr Seth Berkley, the International Aids Vaccine Initiative's president, declares that politics and economics are bigger obstacles to progress than "a scientific barrier". Dr Mary Lou Clements-Mann, a researcher for a rival company's vaccine (and who died in a Swissair plane crash off Nova Scotia last year), shrugs off pessimistic "misperceptions". Dr William Heyward, the CDC's Aids vaccine chief, argues that "only through such trials" as the Bangkok project "will further knowledge be gained".
When visitors drop by, Berman outlines his own paper. It sets out how AidsVax is meant to work. "Many lines of evidence suggest that a strong antibody response to the HIV-1 envelope glycoprotein," he explains, "will be an essential feature of any Aids vaccine." Berman sketches what this means on a board in the conference room, across the corridor from Francis's office. The billions spent on Aids have produced unparalleled insights on HIV, which are the platform on which he builds. The virus infects. The immune system checks it with, among other things, specially-tailored antibodies. But the virus mutates around these adversaries. So the immune system tailors new defences. The virus then mutates and immunity responds. It's like a leapfrog competition. Eventually, the immune system tires of all this leaping, packs up and then it's Aids.
Of all the different parts of HIV, the envelope glycoprotein gp120 is the part that mutates the most. This sits in blobs around the virus, like loose balls of wool, on the tips of protruding spikes. Berman zooms on the moment a blob meets a cell, which is 1m times bigger than the virus. Part of the blob's surface locks onto a receptor (like a data-port where cells get information). The blob then unravels and locks another of its parts onto a second sort of receptor on the cell. This cues the cell to pull the virus inside. Infection is complete.
Here, Berman argues, is where AidsVax helps: by blocking this double-lock connection. Summoned in advance, due to earlier vaccination, antibodies stick to key parts of the blob and so stop it from locking on the cell. If the virus is a burglar, these antibodies are bullterriers, waiting for a leg to appear through the window on which to snap their jaws. Once they've got hold, the virus is paralysed, to be disposed of by other kinds of cell.
He makes things sound simple. Visitors are impressed. Investors wonder: why dither in Bangkok? But the science expounded in the journal issue doesn't convince many people who grasp the detail. "It's a waste of time," Dr Robert Gallo, America's pre-eminent retrovirologist, told me. Prof Andrew McMichael, Aids vaccine chief at Oxford's Institute of Molecular Medicine, said: "I wouldn't have the belief that this will work." And Dr Jean-Paul Levy, head of France's vaccine programme, spat: "It forgets one century of science."
For all the plausibility of the journal's special issue, the most detailed analysis of VaxGen's approach was published in February last year in the Journal of Virology, an even more influential publication. More than 500 people - mostly American gay men - took part in preliminary tests of gp120 in the mid-1990s but experts at seven of America's leading research centres found that, despite the shots, 16 vaccine recipients became infected with HIV. That's more than 3% of those getting vaccine, roughly the same percentage as those on placebo.
Molecular biologists were not surprised, although their critique is extremely technical. What it boils down to is that if HIV leapfrogs the immune system - with all its astounding complexity - it will easily do the same with antibodies induced by an off-the-shelf manufactured product. Inducing antibodies to one B strain, or two E strains, or five, or fifty XYZ strains, is like buying insurance against being hit by cars with specified license plates.
VaxGen's answer is to develop products from strains it claims provide "cross-protection" against others. In Bangkok, for instance, the vaccine is AidsVax B/E, including gp120 clones from one B and one E strain. The B strain was isolated from a six-year-old New Jersey boy in 1984, while the E strain was collected from a soldier in Chiang Mai about nine years ago. The plan is to mix 'n' match vaccines in this way to suit the subtypes in different parts of the world. Berman zooms closer and claims that parts of gp120 stay sufficiently constant between the mutating strains to offer a point of attack. Like all proteins, the blob is made from amino acid molecules, which string together like beads in a necklace to make the loose balls of wool. Each bead is made from one of a possible 20 amino acids. Letters are used to denote these acids: G stands for glycine, for instance, R for arginine and Q for glutamine.
Berman says that the vaccine needs to copy the amino acid sequence at a key point in this string. Near to where gp120 locks onto the cell, there is a loose loop of "wool" - not 100 millionth a cell's size - which biologists call V3. Berman zooms again: to the tip of this loop, a string of just six necklace beads. Here, he argues, is a segment that remains more constant than most and induces antibodies which will stick and stop the double-lock connection with the cell. All it needs is for the vaccine and the virus to have the same acids at the tip of this loop.
Using this argument, Berman deduces that the early tests of gp120 offer hope for the experiment after all. Mostly, volunteers studied for the Journal of Virology were injected with gp120 cloned from the New Jersey strain, in which the necklace in the V3 loop's tip has the beads GPGRAF (meaning: glycine, proline, glycine, arginine, alanine, and phenylalanine). It's a common configuration in North American strains. But Berman argues that some of the volunteers who became HIV-positive despite being vaccinated were infected with strains in which the loop was different: say, GPGRVL (ending with valine and leucine instead). This, he suggests, was why the gp120 didn't protect them. With the commoner strains he believes it did.
At VaxGen's offices, this bottom-line is dazzling. The "special issue" paper quickens pulses. But additional information reveals an oddity, which Berman's presentation overlooks. At the American government's Los Alamos National Laboratory, in New Mexico, staff track amino acid sequences for thousands of HIV strains. And when I asked them to print their data from Thailand, a startling contradiction emerged. The B component in AidsVax B/E - the shots being given to the junkies - has the New Jersey V3 loop tip sequence. It goes: GPGRAF.
According to Berman's argument, the local B strains would need to have the same string of beads. But only 10% of Thai B strains have the New Jersey amino acid sequence. Far more often - in nearly half the strains - there are two different beads in the loop's tip: glutamine (Q) and tryptophan (W). They are GPGQAW. By Berman's own reasoning, the Bangkok junkies are being injected with the wrong vaccine.