Predictive biomarkers for human disease are highly sought after, says In the Pipeline's Derek Lowe. But where, exactly, should one start in the search for biomarkers? A patient's antibodies might be a good place to begin, Lowe says, but "that's some haystack to go rooting around in." Another popular approach — thanks to advanced mass spec and data handling techniques, he says — has been to search patients' tissue samples for "unusual molecules," though that still poses challenges, Lowe says. In a Cell paper, a team of researchers from Florida and Texas takes a different approach. The authors generated a library of several thousand "weirdo 'peptoids'" — antigen molecules that look like "rather weirdly modified peptides," Lowe says — and assembled a microarray and also used them as probes against serum from animal models. And it appears to be effective. In a study of Alzheimer's disease patients, the researchers found three "peptoids" that seemed to be linked to the disease; it's not a solid diagnostic tool, yet, Lowe adds, but these molecules are "pretty impressive."