Skip to main content

Table 2 Effect of read length and base error on HLAminer predictions from targeted assembly of simulated RNA-Seq dataa

From: Derivation of HLA types from shotgun sequence datasets

HLA allele resolution Base error (%) Read length (nucleotides) Sensitivity (mean ± SD%) Specificity (mean ± SD%) Ambiguity(mean ± SD%)
Two-digit 1.0 50 13.62 ± 2.80 92.86 ± 10.10 19.06 ± 16.53
   75 62.32 ± 3.62 90.27 ± 3.28 8.93 ± 2.67
   100 95.72 ± 0.53 96.31 ± 0.02 4.46 ± 3.84
   150 97.97 ± 2.80 95.73 ± 3.63 0.00 ± 0.00
Two-digit 0.5 100 93.04 ± 4.60 96.39 ± 1.44 2.91 ± 1.69
  1.0   95.72 ± 0.53 96.31 ± 0.02 4.46 ± 3.84
  2.0   64.64 ± 4.79 96.13 ± 1.43 13.40 ± 3.02
  3.0   6.67 ± 2.51 100.00 ± 0.00 8.59 ± 8.34
Four-digit 1.0 50 7.78 ± 1.92 60.51 ± 20.02 27.78 ± 4.81
   75 51.94 ± 2.93 77.36 ± 5.44 37.38 ± 11.16
   100 86.84 ± 1.75 88.13 ± 1.41 36.32 ± 4.76
   150 93.33 ± 3.63 93.07 ± 2.91 22.87 ± 2.40
Four-digit 0.5 100 84.72 ± 5.42 89.65 ± 2.25 24.03 ± 3.00
  1.0   86.84 ± 1.75 88.13 ± 1.41 36.32 ± 4.76
  2.0   56.94 ± 1.73 87.49 ± 4.77 39.14 ± 5.73
  3.0   4.44 ± 2.10 68.69 ± 17.03 37.22 ± 25.62
  1. aIn triplicate experiments, 5 million read pairs 50, 75, 100 or 150 nucleotides in length (top) and 100-nucleotide read pairs having 0.5, 1, 2 or 3% errors (bottom) were randomly generated from 20 sets of transcripts, each containing 6 randomly chosen reference HLA alleles. HLAminer predictions derived from targeted read assembly were compared to each reference set and the performance of HLAminer was assessed by measuring the specificity, sensitivity and ambiguity. SD, standard deviation.