Table 2 Summary of FEP+ and Prime statistics in predicting mutational resistance or sensitivity to FDA-approved TKIs

From: Predicting resistance of clinical Abl mutations to targeted kinase inhibitors using alchemical free-energy calculations

Dataset

Method

N quant

MUE (kcal mol−1)

RMSE (kcal mol−1)

N class

Accuracy

Specificity

Sensitivity

all

FEP+

142

\(0.79_{0.67}^{0.92}\)

\(1.07_{0.89}^{1.26}\)

144

\(0.88_{0.82}^{0.93}\)

\(0.94_{0.89}^{0.98}\)

\(0.47_{0.25}^{0.69}\)

all

Prime

142

\(1.14_{0.93}^{1.35}\)

\(1.70_{1.40}^{1.98}\)

144

\(0.73_{0.66}^{0.80}\)

\(0.76_{0.69}^{0.84}\)

\(0.53_{0.30}^{0.76}\)

xtals

FEP+

129

\(0.82_{0.69}^{0.95}\)

\(1.11_{0.91}^{1.30}\)

131

\(0.87_{0.81}^{0.92}\)

\(0.93_{0.88}^{0.97}\)

\(0.50_{0.29}^{0.74}\)

xtals

Prime

129

\(1.16_{0.96}^{1.37}\)

\(1.72_{1.41}^{2.00}\)

131

\(0.72_{0.64}^{0.79}\)

\(0.75_{0.67}^{0.83}\)

\(0.50_{0.25}^{0.73}\)

axitinib

FEP+

26

\(0.70_{0.50}^{0.93}\)

\(0.91_{0.64}^{1.14}\)

26

\(0.85_{0.69}^{0.96}\)

\(0.85_{0.69}^{0.96}\)

NA

axitinib

Prime

26

\(1.05_{0.53}^{1.71}\)

\(1.85_{0.96}^{2.61}\)

26

\(0.88_{0.73}^{1.00}\)

\(0.88_{0.73}^{1.00}\)

NA

bosutinib

FEP+

21

\(0.96_{0.55}^{1.42}\)

\(1.41_{0.77}^{1.97}\)

21

\(0.76_{0.57}^{0.95}\)

\(0.88_{0.71}^{1.00}\)

\(0.25_{0.00}^{1.00}\)

bosutinib

Prime

21

\(1.13_{0.60}^{1.83}\)

\(1.80_{0.92}^{2.62}\)

21

\(0.81_{0.62}^{0.95}\)

\(0.82_{0.62}^{1.00}\)

\(0.75_{0.00}^{1.00}\)

dasatinib

FEP+

20

\(0.76_{0.49}^{1.13}\)

\(1.07_{0.59}^{1.57}\)

21

\(0.90_{0.76}^{1.00}\)

\(0.94_{0.79}^{1.00}\)

\(0.80_{0.33}^{1.00}\)

dasatinib

Prime

20

\(1.05_{0.61}^{1.54}\)

\(1.48_{0.95}^{1.92}\)

21

\(0.86_{0.71}^{1.00}\)

\(0.88_{0.69}^{1.00}\)

\(0.80_{0.33}^{1.00}\)

imatinib

FEP+

20

\(0.82_{0.53}^{1.15}\)

\(1.09_{0.69}^{1.43}\)

21

\(0.86_{0.71}^{1.00}\)

\(1.00_{1.00}^{1.00}\)

\(0.40_{0.00}^{0.83}\)

imatinib

Prime

20

\(1.32_{0.91}^{1.81}\)

\(1.69_{1.15}^{2.26}\)

21

\(0.43_{0.24}^{0.67}\)

\(0.50_{0.25}^{0.75}\)

\(0.20_{0.00}^{0.67}\)

nilotinib

FEP+

21

\(0.82_{0.57}^{1.12}\)

\(1.06_{0.69}^{1.39}\)

21

\(0.86_{0.67}^{1.00}\)

\(0.94_{0.80}^{1.00}\)

\(0.50_{0.00}^{1.00}\)

nilotinib

Prime

21

\(1.50_{1.06}^{1.97}\)

\(1.86_{1.43}^{2.25}\)

21

\(0.48_{0.24}^{0.67}\)

\(0.53_{0.29}^{0.75}\)

\(0.25_{0.00}^{1.00}\)

ponatinib

FEP+

21

\(0.87_{0.62}^{1.16}\)

\(1.09_{0.70}^{1.46}\)

21

\(1.00_{1.00}^{1.00}\)

\(1.00_{1.00}^{1.00}\)

NA

ponatinib

Prime

21

\(0.94_{0.50}^{1.54}\)

\(1.57_{0.69}^{2.44}\)

21

\(0.81_{0.62}^{0.95}\)

\(0.81_{0.62}^{0.95}\)

NA

Glide

FEP+

13

\(0.50_{0.26}^{0.78}\)

\(0.69_{0.35}^{0.97}\)

13

\(0.92_{0.77}^{1.00}\)

\(1.00_{1.00}^{1.00}\)

\(0.00_{0.00}^{0.00}\)

Glide

Prime

13

\(0.91_{0.39}^{1.56}\)

\(1.45_{0.54}^{2.22}\)

13

\(0.85_{0.62}^{1.00}\)

\(0.83_{0.58}^{1.00}\)

\(1.00_{0.00}^{1.00}\)

  1. Accuracy, specificity, and sensitivity were computed to assess two-class prediction performance:
  2. resistant (ΔΔG > 1.36 kcal mol−1) or susceptible (ΔΔG ≤ 1.36 kcal mol−1)
  3. 95% CIs (sub-/superscripts) were estimated from 1000 bootstrap replicates. The sensitivity for axitinib and ponatinib is NA, because there is no resistant mutation for these two drugs
  4. Nquant Number of mutations for which quantitative metrics were evaluated, Nclass number mutations for which classification metrics were evaluated, All all mutations, xtals all mutations for which co-crystal structures were available, Glide erlotinib and gefitinib