Distribution of low quality bases along the PhiX reference genome. Analysis was performed on reads derived from an Illumina PhiX library (PhiX-95nt data set). (a) Number of bases within B-tails (consecutive bases of Q-score = 2 at the 3' end of a read) per position. (b) Average Q-score of bases in untrimmed reads. (c) Average Q-score of bases in B-tail trimmed reads. (d) Observed per-base substitution error rate. Calculations for (a-d) were performed separately for the forward strand (green) and reverse strand (red). Low quality values accumulated in certain regions even after removal of B-tails. The peaks of observed error rates occur at positions where increased low quality counts are detected, and in most cases the peak is seen only on one strand.

Observed error rates of 2 × 95-nucleotide HiSeq reads by cycle (averaged across all flow cell tiles). Read 1 (left) and read 2 (right) were analyzed separately for PhiX-95nt data (a) and Bv-95nt data (b). PhiX and sugar beet DNA was sequenced in the same lane, and reads were mapped against the PhiX or ZR reference sequences, respectively.

Frequencies and context of sequencing errors and quality scores compared to observed error rates. The sugar beet sample (yellow) and the Arabidopsis sample (blue) were each sequenced together with PhiX DNA (red and green, respectively) on a HiSeq2000 sequencing instrument. PhiX DNA only (black) was sequenced on a GAIIx. (a) Sequence context of substitution errors. The frequency of neighboring bases one position upstream and downstream of an error position is displayed. Sequence triplets were summarized for all types of base substitutions at the central position (indicated by an 'e'). We counted reads spanning the triplet positions and ignored potential further substitution errors within the triplet sequence of the read. The frequency was determined by dividing the occurrence of a triplet containing a central substitution error by the occurrence of all triplets with the same marginal bases but variable central base. The display of triplets is ordered by increasing average frequency in the HiSeq data. (b) Frequency of base substitution errors. For each sample, the proportion of each substitution is indicated (ordered by increasing average frequency in the HiSeq samples). (c) Rates of insertions or deletions in homopolymer tracts normalized by homopolymer length. Homopolymers longer than seven bases were present only in the two plant samples. Homopolymers of length 16 to 19 in the Bv-95nt data and of length 26 to 29 in the At-100nt data were each covered by less than 50 reads. (d) Expected versus observed error rates. Expected error rates according to quality scores (Q) were calculated for Q = 2 to Q = 40 (solid diagonal line). For each sample the uniquely aligned bases were grouped by quality score, and the observed error rate was determined from the number of observed substitution errors for each Q separately.

Distribution of read coverage depth for (a) Bv-95nt reads and (b) Phix-95nt reads. Read coverage was computed per base. In three separate calculations we considered all positions (black), positions in regions below (red) and positions in regions above (blue) the average GC content (%GC) of the reference. The regional %GC was determined based on a window of 250 bases upstream and 250 bases downstream of each position. In contrast to PhiX (b) the coverage variation in the sugar beet sample (a) is related to the %GC.

Alignment of reads before (a) and after (b) B-tail trimming in a selected region of the ZR reference (positions 63,633 to 63,662). Uniquely mapped reads from the Bv-95nt data set were visualized using the Tablet browser [17]. Forward matching reads are shown in grey, reverse matching reads are shown in blue, mismatch bases are shown in white. Long white stretches are uncalled bases. Mismatches accumulated in one region, and almost all mismatches are eliminated after B-tail removal.

Mean quality scores of correct and wrong bases sequenced (PhiX-95nt data) at error-prone positions and all other positions in the PhiX reference. The bases covering 161 positions of significantly elevated error rates (A, B) in the PhiX reference show lower average quality scores compared to the bases covering other positions (C, D). This is true for correctly called bases (A, C) as well as for incorrectly called bases (B, D).