Removing adapters for assembly

Question

[Ric]

Hi, I ran fastp with this parameters fastp --detect_adapter_for_pe --correction ... and it removed the adapters and trimmed the paired-end reads. Which read length should I remove form the below read length distribution?

   #count #lenght
   329 15
    552 16
    503 17
    736 18
    769 19
   1026 20
    972 21
   1291 22
   1380 23
   1767 24
   1599 25
   1873 26
   1915 27
   2088 28
   2086 29
   2233 30
  55851 31
  52576 32
  51137 33
  49592 34
  47922 35
  46953 36
  45391 37
  44361 38
  44025 39
  43604 40
  43689 41
  42873 42
  42939 43
  42076 44
  41975 45
  42687 46
  41848 47
  42640 48
  42421 49
  42740 50
  43838 51
  44403 52
  45416 53
  46181 54
  47300 55
  48651 56
  49389 57
  50291 58
  51105 59
  52482 60
  53037 61
  53665 62
  57750 63
  58925 64
  59487 65
  60286 66
  61682 67
  61284 68
  62572 69
  63134 70
  63262 71
  63852 72
  64578 73
  64873 74
  64994 75
  65885 76
  66335 77
  67944 78
  68972 79
  69102 80
  69150 81
  69667 82
  70801 83
  72306 84
  72673 85
  74276 86
  75306 87
  75331 88
  77290 89
  78503 90
  83206 91
  86050 92
  86968 93
  88749 94
  90323 95
  91604 96
  94306 97
  97176 98
  99139 99
 102145 100
 104107 101
 106918 102
 108874 103
 112628 104
 116593 105
 120706 106
 124080 107
 127632 108
 132103 109
 136848 110
 141106 111
 147163 112
 150847 113
 157369 114
 162760 115
 168054 116
 174119 117
 179892 118
 186705 119
 193111 120
 200338 121
 207203 122
 214064 123
 222115 124
 230228 125
 238055 126
 246812 127
 256007 128
 264440 129
 273321 130
 283574 131
 291681 132
 301680 133
 311049 134
 321920 135
 331860 136
 339637 137
 350607 138
 363766 139
 378285 140
 395352 141
 440685 142
 532445 143
 427897 144
 485367 145
 881886 146
 449135 147
 461934 148
 471502 149
 481966 150
514886614 151

Answer 1

[Mensur Dlakic]

It depends what you are doing with the data afterwards. If the data is meant for a de novo assembly, I usually don't throw away anything. My understanding is that the assembler will automatically handle this and will discard the reads shorter than the k-mer overlap value. It may make some difference if you are using it for mapping, though probably not much. If you make a cut-off at 75, you'd be throwing away <0.5% of reads and <0.2% of sequence data. It would still be less than 1% of reads even at 100.