Table 1.

Symmetric and asymmetric depth-dependent potentials derived from experimental data or statistics.

author	year	α/β a	E/S b	S/A c	ref	comments	#prot d	embedding e	SASA f	bin size	functional form
Hessa & von Heijne, biological scale	2007	α	exp	sym	22	Inserted a TM segment into the leader peptidase protein, into which all amino acids were introduced at different depths. Translocation into microsomes was quantified by the number of glycosylation sites on either terminus of the TM segment. 324 19-residue TM segments were measured to compute ΔΔG’s	1	center of TM segment is at membrane center	protein is 3-helix bundle, so most are lipid-exposed	residue	double Gaussian for WY, single Gaussian for others
Elazar & Fleishman, dsTbL	2016	α	exp	asym	12	Combined sequence libraries with TOXCAT assay in whole cells. Measured TM span expression, insertion and association depending on residue depth with orthogonal antibiotic resistance markers. 472 mutants were tested 100 times each in a high-throughput manner to compute ΔΔG’s	1	membrane center was estimated by aligning ILMF profiles’ troughs	single helix, so all lipid-exposed	residue	4 D polynomials
Senes & DeGrado, Ez potential	2007	α	stat	sym	23	statistical potential derived from 24 MPs in a symmetric manner; insufficient counts for C	24	protein COM at membrane center	no distinction	2 Å	double Gaussian for WY, sym sigmoidal for others
Ulmschneider, implicit membrane potential	2005	α	stat	asym	33	statistical potential derived from 46 MPs in an asymmetric manner; resolution ≤4 Å; insufficient counts for CST;	46	centered DSSP TM spans at membrane center	SASA probe radius 1.4 Å	2 Å	double Gaussian for RKDEHWY, single Gaussian else
Schramm & DeGrado, Ez potential	2012	α	stat	asym	25	statistical potential derived from 76 MPs in an asymmetric manner; sequence similarity ≤30%, resolution ≤ 3.5 Å	76	OPM embedding	SASA probe radius 1.9 Å	2 Å	sigmoid, Gaussian or combination of the two
this work	2018	α	stat	asym	this	statistical potential derived from 239 MPs in an asymmetric manner; sequence similarity ≤30%, resolution ≤3 Å	239	topology from OPM but embedding from PDBTM	lipid-exposed vs buried35	3 Å	double Gaussian
Moon & Fleming, sidechain hydrophobicity scale	2011	β	exp	sym	20	reversible GnHCl (un)folding of OmpLA into DLPC vesicles to derive symmetric profile; an A residue at the membrane center was mutated into all 19 other amino acids; 3 titrations for WT and 2 titrations for mutants to compute ΔΔG’s; potential derived for LR	1	membrane center set halfway between aromatic girdles	only lipid-exposed residues	residue	single Gaussian for LR, no fit parameters given
MacDonald & Fleming, sidechain hydrophobicity scale	2016	β	exp	sym	21	reversible GnHCl (un)folding of OmpLA into DLPC vesicles; residues at different depths were mutated into WYF; 3 titrations for WT and 2 titrations for mutants to compute ΔΔG’s;	1	center from MD simulations: COM of the protein and phosphate atoms	only lipid-exposed residues	residue	linear for WYF
Hsieh & Nanda, Ez potential	2012	β	stat	sym	26	statistical potential derived from 35 MPs in a symmetric manner; sequence similarity ≤26%; insufficient counts for CM	35	embedding from OPM TM spans	SASA > 0.2	3 Å	double Gaussian for WYFG, sym sigmoid for others
Wimley	2002	β	stat	asym	54	statistical 3-state hydrophobicity scale from 15 non-redundant β-barrels;	15	aromatic girdles were used	lipid vs water exposed	regions	no fitting done
Jackups & Liang, positive outside rule	2006	β	stat	asym	27	sequence similarity ≤ 26%; resolution ≤2.6 Å; derived statistics for regions, depending on z and burial, but no depth-dependent potential like the others, they use it to derive a basic energy function for barrel prediction based on H-bonds	19	embedding from OPM TM spans	regions	regions	no fitting done
Slusky & Dunbrack, charge outside rule	2013	β	stat	asym	15	statistical potential derived from 55 MPs in an asymmetric manner; sequence similarity ≤50%; resolution ≤3.5 Å; only averages of AA groups were fit, but not individual AA types; insufficient counts for PCMT	55	N/C termini are inside, membrane center defined where phospholipid meets LPS and aromatic girdle set to −12Å	lipid vs water exposed	3 Å	RKDE to P2, NQHS to P2, AGILV to P2, FWY to P4, no fitting parameters given
Lin & Liang, TMSIP	2017	β	stat	asym	28	19 MPs were used to derive an energy function that includes a membrane burial term and inter- and intra-strand H-bond interaction energies27, the energy function was used to derive a statistical potential for ΔΔG’s tested on 24 MPs; sequence similarity ≤26%; resolution ≤2.6 Å	19	embedding from OPM TM spans	lipid vs water exposed	residue (regions for deri- vation)	double Gaussian for WY, single Gaussian for others
this work	2018	β	stat	asym	this	statistical potential derived from 96 MPs in an asymmetric manner; sequence similarity ≤50%, resolution ≤3 Å; insufficient counts for C	96	topology from OPM but embedding from PDBTM	lipid-exposed vs buried35	3 Å	4D polynomial for IM, double Gaussian for others

^aα-helical or β-barrel.

^bExperimental or statistical.

^cSymmetric or asymmetric.

^dnumber of proteins used for derivation.

^eHow proteins were embedded/centered in the membrane.

^fwas distinction made between lipid-exposed and lipid-buried residues or how was SASA calculated.