A method and an apparatus for genome assembly are provided. The method comprises: filtering a short-fragment-sequence output from end sequencing of an large insert-size library to remove unqualified sequence; aligning the filtered short-fragment-sequence onto a reference genome sequence, wherein, the filtered short-fragment-sequences comprise paired short-fragment-sequences; sorting the paired short-fragment-sequence after alignment into soap reads sequence, single reads sequence and unmap reads sequence based on the aligning result, and counting the number of each sort of sequence; calculating a distance between the paired soap reads on a fragment of the reference genome sequence, wherein a pair of the paired soap reads can be aligned onto a same fragment of the reference genome sequence; and counting a distance distribution of each pair of soap reads on the reference genome sequence; and assembling the genome sequence by using the paired single reads upon the distance distribution meeting a requirement of a threshold, wherein a pair of the paired single reads can be aligned onto two different fragments of the reference genome sequence.