Abstract: In order to analyze the relationships between chemical components and sensory quality of flue-cured tobacco leaves, cigarette samples made from 40 different tobacco leaf samples from major domestic flue-cured tobacco production areas with the same primary processing parameters and manufacturing parameters were taken as the research objects. A Lasso regression model was established to screen and determine the chemical components in the leaf and in mainstream cigarette smoke. The results obtained from the Lasso regression model, PCA principal component regression model, and partial least squares regression model were compared. The classical linear model was used as the basis to study the relationships between the contents of 90 chemical components in mainstream smoke and the main sensory indexes. The results showed that: 1)The Lasso regression model was more effective in reducing the number of independent variable factors. 2)Strength and flavor style in the sensory indexes were significantly linearly correlated with some chemical components with the verification coefficients of determination of 0.70 and 0.84, respectively. The prediction accuracy was higher, but the linear correlations between the other sensory indexes and the chemical components were not significant. 3)According to the linear regression model, nicotine, butyric acid, and 2-ethylpyridine were positively correlated with strength, while 3-methyl-2-cyclopenten-1-one was negatively correlated with strength (P < 0.05); nicotyrine, nicotine, norsolanadione, 3-hydroxy-β-dihydrodamascone, 2-ethylpyridine, and methylcyclopentenolone were positively correlated with the fresh flavor style(P < 0.05), as were 2-methylpyridine, 3-oxo-α-ionol, neophytadiene, and furanone with the robust flavor style(P < 0.05).