Table 1.

Changes in experimental binding and DSC4.1 computed electrostatic energy term (ΔΔE_elec) for interaction of cetuximab (C225), necitumumab (11F8) and panitumumab (pani) with wild type and epitope mutated EGFR variants.

Mutationa	KDmut/KDWT b			Experimental Bindingc			ΔΔEelec
	C225	11F8	pani	C225	11F8	pani	C225	11F8	pani
S440L	>1000	8.8 ± 1.2	NB	⊗	●	⊗	0.58	0.03	0.12
G441E	NB	NB	NB	⊗	⊗	⊗	4.58	3.80	0.73
G441R	NB	NB	NB	⊗	⊗	⊗	1.30	−0.06	0.00
K443T	120 ± 20	2.3 ± 0.1	✓	⊗	●	●	0.80	−0.06	0.21
I467M	46 ± 4	3.9 ± 0.4	NB	⊗	●	⊗	0.09	−0.13	0.08
S468R	>1000	2.0 ± 0.2	3.4	⊗	●	●	1.64	−0.51	−0.08
Q384A	2.1 ± 1.1	13 ± 1	–	●	⊗	–	−0.04	0.08	–
K443A	1.3 ± 0.4	0.9 ± 0.1	−	●	●	−	0.67	−0.38	–
S468I	0.7 ± 0.1	0.27 ± 0.03	–	●	●	–	0.28	−0.38	–
N473A	0.9 ± 0.1	0.6 ± 0.1	–	●	●	–	−0.02	−0.20	–
				TPd			2	6	1
				TN			6	2	4
				FP			0	1	0
				FN			2	1	1
				Accuracy			81 %
				Precision			90 %

^aThe first 6 variants (bold) are associated with cetuximab and/or panitumumab resistance (Supplementary Table S1), lower 4 mutations from Li et al (22).

^bK_D values for cetuximab (C225) and necitumumab (11F8) are from Supplementary Table S2 (S440L, K443T, I467M, S468R) and from (22) (Q384A, K443A, S468I, N472A). Data for panitumumab (pani) for S468R is from (23). NB – no binding from (14, 17) and Octet data. ✓ indicates binding but no affinity reported (14).

^c⊗ No binding (K_D value more than 10-fold greater than wild type).

● WT binding (K_D value within 10-fold of value for wild type).

^dFor counting of true vs false predictions, an EGFR variant is considered an experimental “binder” if the K_D value is less than 10-fold greater than the K_D value for wild type (●). The computational filter is set at ΔΔE < 0. True positives (TP, binder (●) with negative computed energy) and true negatives (TN, non-binders (⊗) with positive or zero computed energy term) are highlighted in bold. Accuracy is defined as (TP + TN)/(Total number of variants) and precision is defined as TP/(TP + FP).