The possibility of using folate-producing lactic acid bacteria as an adjunct culture in wheat sourdough and the effect of different sourdough formulations on the measurable folate content in wheat bread were investigated. Among 43 strains isolated from traditional Khorasan dairy products, three strains with the highest extracellular folate production were selected. Four sourdough treatments including S1 (flour, water, yeast and selected LAB), S2 (flour, water and selected LAB), S3 (flour, water and yeast) and S4 (control; flour and water) were prepared and the corresponding breads (B1–B4) were produced. Folate content in sourdough and bread was measured using high-performance liquid chromatography. The pH and acidity of the samples and the changes in pH and acidity during fermentation were measured. Changes in folate content during the bread process were investigated and the percentage of folate loss was determined. In addition to classical statistical tests, principal component analysis and partial least squares regression were used as multivariate exploratory tools to examine patterns of variation in the data. The results showed that the highest folate content was observed in sourdough S1 (14.04 µg/100 g) and B1 bread (13.32 µg/100 g) prepared with sourdough containing yeast and selected LAB, and the lowest content was observed in S4. The simultaneous presence of yeast and LAB increased folate production during fermentation and improved the final folate level of bread. PCA results showed that the two principal components explained more than 94% of the variance in the data and described the pattern of variation related to pH, acidity, and folate in the samples. Two PLSR approaches were examined based on final fermentation characteristics and dynamic changes during fermentation. The model based on dynamic changes showed a better fit than the model based on final characteristics; However, due to the limited number of samples and the lack of independent validation, the PLSR results can only be interpreted at an exploratory level and further studies are needed to develop valid predictive models.