Metagenome sequencing is a key technology for studying microbiome.A single metagenome sample usually contains millions of short reads from diverse species,with different genome length and abundance.The high similarity between different organism genomes along with the sequence bias will cause uneven coverage in a single genome.We investigated the uneven coverage in mock community data as well as really metagenome data and got some interesting observations.Experiments showed that the coverage bias will have influence on the downstream analysis such as using partial genome coverage to predict the genome length1.A probabilistic method was also developed to model and correct the coverage bias.