We evaluated a microelectromechanical systems (MEMS) microsensor array with temperature-controlled chemi-resistive elements for use as a noninvasive clinical diagnostic tool to detect the presence or absence of trace amounts of disease biomarkers in simulated breath samples. The microsensor environment was periodically altered between air (78% N2, 21% O2 by volume, 20% relative humidity) and simulated breath (77% N2, 16% O2, 4% CO2 by volume, 80% relative humidity) samples creating a dynamic background. Acetone, a disease marker for diabetes, was spiked into select simulated breath samples at relevant concentrations ( 0.5 Â¿mol/mol to 8 Â¿mol/mol) to pose a diagnostic problem for the sensor array. Using standard statistical dimensionality reduction and classification algorithms, we compared the ability of a variety of sensing materials to detect and recognize the disease marker. Our analyses indicate that the porous, doped nanoparticle materials (Sb:SnO2 microshell films and Nb :TiO2 nanoparticle films) are best for the recognition problem (acetone present versus absent), but that WO3 and SnO2 films are better at the quantification task (high versus low concentrations of acetone).