Skip to main content
. 2022 Feb 8;14(1):2020203. doi: 10.1080/19420862.2021.2020203

Table 2.

Evaluation of humanization methods. Separately for each column, values above 75% percentile are marked in bold, values below 25% percentile are marked in italics. Humanness change was computed as average absolute difference of OASis medium identity (or T20 score) of the humanized sequence and the parent sequence. Therefore, an increase of +34% refers to the absolute change in the humanness score (e.g. from 40% to 74%), not a relative change. Parental preservation was calculated as sequence identity of the parental and humanized sequence under Kabat numbering, in full sequence or Vernier regions only. Humanizing mutation precision was calculated as number of shared mutations (made both in predicted sequence and in experimentally humanized sequence), divided by total number of mutations made in the predicted sequence

 
 
 
Humanness improvement
Preservation
Humanizing mutation precision
    Method OASis T20 Total Vernier Total Vernier
25 known pairs Experimental +34% +13% 80% 86% - -
Sapiens*1 +30% +10% 89% 90% 77% 49%
Sapiens*2 +33% +12% 86% 87% 74% 42%
Sapiens*3 +34% +12% 86% 86% 73% 40%
Sapiens*4 +34% +13% 86% 86% 72% 40%
Hu-mAb +14% +6% 89% 91% 72% 54%
Straight graft +36% +14% 83% 80% 65% 35%
Vernier graft +32% +11% 85% 100% 70% -
  Experimental +30% +11% 84% 87% - -
152 putative pairs Automatic
germline
Sapiens*1 +30% +10% 89% 89% 72% 48%
Hu-mAb +4% −1% 93% 94% 36% 25%
Straight graft +33% +13% 84% 80% 64% 36%
Vernier graft +29% +11% 87% 100% 70% -
Manual
germline
Sapiens *1 +34% +13% 83% 83% 82% 49%
Hu-mAb +15% +3% 90% 93% 65% 48%
Straight graft +35% +13% 82% 76% 81% 45%
Vernier graft +30% +11% 85% 100% 89% -