Background: DNA from the host, and extraction biases can significantly interfere with microbiota assessment and increase the cost of high throughput sequencing. This study aimed to find a better method to enable a higher concentration of microbial DNA to be extracted from the bronchoalveolar lavage fluid (BALF) samples with a higher proportion of host cells. Methods: Two subjects were enrolled in this study. DNA was extracted from 3.0 ml of the fresh or frozen BALF samples of each subject using QIAamp DNA Microbiome Kit. DNA libraries were constructed according to the manufacturer's instructions (illumina). Cluster generation, template hybridization, isothermal amplification, linearization, blocking, denaturing, and hybridization of the sequencing primers were performed. Raw data was uploaded to MG-RAST v3 and analyzed. Results: Through preliminary analysis of sequence characteristics, it was found that the ratio of human genome in clean reads of each DNA sample from frozen BALF samples was significantly lower than those from fresh BALF samples (Patient 1:59.80% vs. 68.60%, Patient 2: 47.91% vs. 66.89%, respectively). For each individual, numbers of hits from frozen samples by same database were obviously higher than those from fresh samples. Tiny difference on the top 50 most abundant classified phyla was observed between fresh BALF sample and frozen BALF sample from the same individual. The different classified phyla belonged to low abundance phyla of Eukaryota. No significant difference on the principal bacterial community structure between fresh BALF samples and frozen BALF samples by comparing the bacterial community structure of fresh BALF samples with frozen BALF samples. Conclusion: Cryopreservation of BALF samples enables conservation of a higher yield of microbial DNA from samples and a higher fraction of host cells. It will not affect the principal structure of bacterial community.