Neural network algorithm enables mass calibration autocorrection for miniature mass spectrometry systems

Mass spectrometry (MS) is a powerful analytical technology widely used in a broad range of applications. Laboratory-scale mass spectrometers, however, are hardly used outside the analytical laboratories due to the large sizes and weights. Miniature mass spectrometers are therefore developed to facilitate on-site MS analysis. How to stabilize their analytical performances under complex environmental conditions on-site is a challenging problem, which needs to be addressed for the development of miniature MS instrumentation. Here, we report a neural network algorithm which enables automatic mass calibration corrections for a Cell miniature MS system (PURSPEC Technologies Inc.). To simulate the change of complex environmental conditions on-site, variations of temperature from 5℃ to 40℃, pressure from 98647 Pa to 99406 Pa, humidity from 30 % to 65 %, were employed. The mass accuracy, characterized by the difference between measured mass and nominal mass, after autocorrection of the algorithm was within 0.08 Da.

Miniature mass spectrometer is vulnerable to environmental changes, such as temperatures, relative humidity, and air pressure to produce mass shift, which affects the mass accuracy of the instrument for on-site analysis. To address this practical problem, we establish a mass shift correction algorithm based on a neural network algorithm for miniature MS systems. After correction, the neural network algorithm showed a high mass accuracy within 0.08 Da. This is largely attributed to that the neural network had more internal connections with the samples to automatically convert lower-level features into higher-level features. The neural network algorithm developed here would be useful for on-site MS analysis using miniature MS systems.