1. Accurate Sample Assignment in a Multiplexed, Ultrasensitive, High-Throughput Sequencing Assay for Minimal Residual Disease 
Jack Bartram, Edward Mountjoy, Tony Brooks, Jeremy Hancock, Helen Williamson, Gary Wright, John Moppett, Nick Goulden, Mike Hubank 
The Journal of Molecular Diagnostics 
Volume 18, Issue 4, Pages (July 2016)
DOI: /j.jmoldx
Copyright © Terms and Conditions

2. Figure 1
Nested PCR and final library product for IGH VDJ (variable, diversity, joining) amplicon sequencing on the Illumina MiSeq (Illumina Inc., San Diego, CA). A: Family variable heavy chain segment (VH) and joining heavy chain segment (JH) primers with complementary partial index sequences are used for amplify the VDJ junction of the rearranged IGH gene, containing the hypervariable region (purple). B: In the second stage, index sequences (blue) and Illumina P5 and P7 platform adaptors sequences are added. C: The final amplicon construct with sequencing strategy. One or two index reads were used, depending on indexing strategy, with single-end sequencing read from the JH (P5) end of the amplicon.
The Journal of Molecular Diagnostics  , DOI: ( /j.jmoldx )
Copyright © Terms and Conditions

3. Figure 2
Misassignment of sequences from sequencing run A7BK7. Indexed reads corresponding to samples are shown in red. Reads assigned to indices not included in the sequencing run are shown in blue. Eight–base pair indices used as per Kozarewa et al.33
The Journal of Molecular Diagnostics  , DOI: ( /j.jmoldx )
Copyright © Terms and Conditions

4. Figure 3
Comparison of quality statistics between reads assigned to real samples (true) and reads misassigned to indices not included in the run (examples using data from run A7ELC). A–C: Base quality of sequencing reads shown by position (A), mean quality score distribution for all sequences (B), and index read quality (C) from a representative true indexed sample. D–F: Base quality of sequencing reads shown by position (D), mean quality score distribution for all sequences (E), and index read quality (F) from a representative misassigned sample. Pooled libraries were sequenced (150 bases, single-end) on an Illumina MiSeq (Illumina Inc., San Diego, CA). Data are expressed as the median value (middle red line), with the yellow box representing the inter-quartile range (25-75%), the top and bottom whiskers representing the 10% and 90% points, and the blue line representing the mean quality.
The Journal of Molecular Diagnostics  , DOI: ( /j.jmoldx )
Copyright © Terms and Conditions

5. Figure 4
Accurate quantification of minimal residual disease down to 1 in 1 million cells, using a spike in control quantified by digital PCR. Three separate serial dilutions of cell lines; SUPB15 (A), TOM-1 (B), and REH (C), spiked into 1 million cells equivalent to pooled normal lymphocyte DNA. Samples sequenced on Illumina MiSeq (Illumina Inc., San Diego, CA) (>2 million reads per sample), sequenced using the optimized HTS-MRD strategy developed in this article and demultiplexed using Error-Aware Demultiplexer.
The Journal of Molecular Diagnostics  , DOI: ( /j.jmoldx )
Copyright © Terms and Conditions

6. Figure 5
Minimal residual disease (MRD) analysis performed in five patients at the end of induction chemotherapy for childhood B-lineage acute lymphoblastic leukemia (day 28 MRD). The current gold standard ASO–real-time quantitative PCR (ASO-qPCR) technique is compared to a standard high-throughput sequencing (HTS) work flow (A) and an optimized workflow using Error-Aware Demultiplexer (B). The sequencing run included a diluted (10−1) diagnostic sample for each patient to simulate an increased probability of misassignment. MRD samples were also spiked with cell line DNA for quantification purposes. Errors introduced during standard multiplexed sequencing result in misdiagnosis in patients N001 and N005 due to misassignment (A). These incorrect calls due to misassignment are removed (B) using the workflow presented.
The Journal of Molecular Diagnostics  , DOI: ( /j.jmoldx )
Copyright © Terms and Conditions