Objective To study and establish a new molecular diagnostic model of saliva protein fingerprints in gastic cancer patients with spleen deficiency. Methods Saliva samples of 57 gastric cancer patients with no surgery and chemoradiotherapy（ 40 cases of spleen deficiency

17 non-spleen deficiency） and 45 healthy volunteers were collected and tested by weak cation exchange（ WCX） magnetic nanoparticles combined with matrix-assisted laser desorption / ionization time of flight mass spectrometry（ MALDI-TOF MS） to get the protein fingerprints of each group. Then the data were compared to figure out the differentially expressed protein peaks with statistical significance between gastric cancer spleen deficiency group and normal control group as well as that between gastric cancer spleen deficiency group and gastric cancer non-spleen deficiency group. The discriminant value of the variation of several variables

including mass-to-charge ratios（ m / z）

to two group samples were calculated by decision tree algorithm

and the optimal diagnostic model was determined. Results There were 106 differentially expressed protein peaks with statistical significance in the saliva protein mass spectrum comparation between gastric cancer spleen deficiency group and normal control group

and 6 between gastric cancer spleen deficiency group and gastric cancer non-spleen-deficiency group（ P < 0. 05 or P < 0. 01）. Three main different protein peaks with m / z 5439. 67

2411. 67

3619. 18 were screened to establish the diagnostic model of gastric cancer spleen deficiency. The sensibility and specificity of the model are 100% and 93% respectively in clinical regression tests

which are further validated to be 87% and 89% in cross-validation. Conclusions Based on WCX magnetic nanoparticles combined with MALDI-TOF MS technology

the non-invasive saliva protein molecular diagnostic model of gastric cancer spleen deficiency has high sensibility and specificity

and it seems to have excellent efficiency to diagnose gastric cancer spleen deficiency.