(a) Averaged time traces for flows 3, 7, 15, 23 that correspond to 0, 1, 2 and 3-mer ‘A’ flows. (b) Histograms of all 11 ‘A’ flows from a 44 flow sequence. Histograms are representative of homopolymer repeats. K-means clustering method was used to discriminate clusters and perform base calling on homopolymer repeats. When MaxAmp/SSAmp and integrated counts are binned, three-dimensional histograms for the two axes are used to demonstrate the additional discrimination power for homopolymer discrimination of G, A, C and T flows. (c) Three-dimensional (3D) histograms of G homopolymer repeats from flows 5, 25, 29 and 37, which correspond to homopolymer repeats of 0, 1, 2 and 3-mers, respectively. (d) 3D histograms of T homopolymer repeats from flows 4, 8, 12 and 20, which correspond to homopolymer repeats of 1, 0, 5 and 2-mers, respectively. (e) 3D histograms of A homopolymer repeats from flows 3, 7, 15 and 23, which correspond to homopolymer repeats of 0, 1, 2 and 3-mers, respectively. (f) 3D histograms of C homopolymer repeats from flows 6 and 14, which correspond to homopolymer repeats of 0 and 1-mers, respectively. (g) After 44 flows, a random ROI with 357 clusters was chosen and base calling was performed as described in the text. Of the 357 clusters, 336 clusters were successfully aligned, which accounts for 91% of the clusters. Of these 91% aligned reads, the error distributions are shown for the following minimum overlap lengths: 15 bp (blue bars), 20 bp (red bars) and 28 bp (green bars).