A new startup hopes to spin drug candidate straw into medicinal gold by scouring data from failed clinical trials of Alzheimer’s treatments and identifying patient subtypes for whom the drugs may actually work. Alzheon, a Boston-area company, was founded with the belief that recent gains in the scientific community’s understanding of the disease, enabled by new technologies for genetic analysis and neuroimaging, can shine light into a vast morass of failed clinical trial data.

Alzheon will first take a closer look at a drug designed to target the brain-damaging amyloid plaques that are the hallmarks of Alzheimer’s. In the hands of another drug company, the compound had previously passed safety testing but did not show enough of an effect on cognitive impairment in studies of Alzheimer’s patients. However, later analyses suggest that in patients with a certain genetic marker for Alzheimer’s, the drug did slow brain atrophy significantly. Alzheon believes that a follow-up trial will prove the drug to be successful for patients with these markers.

Currently, there are no treatments for Alzheimer’s disease beyond drugs that temporarily alleviate symptoms. This is despite millions of dollars spent and decades of research by academics and drug companies, which seem to announce drug failures with depressing regularity (see “A New Setback for Alzheimer’s Drugs” and “Another Bust for Alzheimer’s Drugs”). Meanwhile, pressure to discover a viable treatment grows as the population ages (see “The Dementia Plague”).

Developing drugs for Alzheimer’s is particularly tricky. Scientists do not know the origin of the disease, so experimental treatments may not be off-target. Alzheimer’s can also look like other forms of dementia, so drugs are tested on a mix of patients, some of whom do not actually have Alzheimer’s. And there is even heterogeneity within Alzheimer’s patients, meaning different patient types need different drugs.

But, while many mysteries remain, the scientific community has made progress in understanding Alzheimer’s. Neuroimaging tools can detect atrophy of different regions of the brain and can also detect the presence of amyloid plaques (see “A Brain Scan for Alzheimer’s”). Genetic research has also identified different inherited DNA variants that make some people more at risk for the disease.

In the past, large pharmaceutical companies have tried to find treatments that work for everybody, and as a result they ran large, inclusive trials, says Alzheon CEO Martin Tolar. A better approach could be to look at smaller patient groups that share risk factors or certain details of disease biology in common. Alzheon plans to do this with existing data, combing through how patients were diagnosed, how they were monitored for improvement, and whether there were signs of the drug working in certain subpopulations. “To our advantage, a lot of data has been generated,” says Tolar, a neurologist who previously worked for the genome analysis company Knome (see “Knome Software Makes Sense of the Genome”).

Sometimes, the failures announced by large drug companies can still point to groups on which a drug may work. When Eli Lilly announced in August last year that its Alzheimer’s drug candidate had failed, the pharmaceutical giant suggested that certain patients—those at an early stage of the disease—may have seen some benefit. Eli Lilly is now recruiting patients with mild Alzheimer’s to test the drug. Other research groups focus on getting the most out of drugs by altering the timing of treatment, including an effort to prevent the disease before any clinical symptoms appear.

A growing body of information suggests that targeting drugs to stratified groups of patients could be a positive step forward, says Martin Farlow, a neurologist who specializes in dementia at Indiana University Health. “The thing that makes me much more optimistic is that over the last five to 10 years there’s been a tremendous amount learned about Alzheimer’s,” he says, including on protein biomarkers that can be found in the cerebral spinal fluid, through brain imaging, and with better understanding the genetics of the illness. Last Friday, the International Genomics of Alzheimer’s Project announced that it had discovered 11 new genes associated with the disease, doubling the previous count. More genetic linkages will help researchers and doctors better “box” patients into different groups, says Farlow.

Alzheon’s lead candidate is a compound that had failed in late-stage clinical trials because no clear effect was seen across the entire patient population. However, in a reanalysis of the trial data, the drug candidate seemed to slow atrophy in certain brain regions.

Tolar and colleagues reanalyzed data from the trial with an eye toward breaking patients into different groups based on genetics and other factors, including identifying patients who carried the strongest genetic risk marker for the disease, the Apo e4 gene variant. They found that in patients with a copy of this variant, the compound slowed atrophy significantly in the hippocampus, a region of the brain particularly susceptible to the damaging effects of Alzheimer’s.

If Alzheon can prove that the compound does indeed slow cognitive decline and brain atrophy in patients with that genetic marker, the company would have a drug that works for a large number of Alzheimer’s patients. “Almost 60 percent of patients with Alzheimer’s carry Apo e4,” says Tolar. Alzheon will begin a study in 2014 to explore the efficacy of the drug in patients with the genetic marker and who exhibit early signs of Alzheimer’s.

Even if this new approach works well, treating Alzheimer’s will likely remain difficult. One issue is that all drugs being tested are pursuing just one theory as to the cause of the disease, when many researchers would argue that other biological changes are the real culprit. “It’s by no means clear that even stratified trials will work,” says Dominic Holland, an Alzheimer’s researcher and neuroimaging specialist at the University of California, San Diego.

That said, given that there are no treatments that work and Alzheimer’s is likely to reach epidemic proportions, stratified trials of existing drugs is worthwhile. “This is a really scary area,” says Holland. “There is a train wreck coming.”