Table 2.

Listing of implied assumptions to formulate valid working hypotheses for the molecular modeling approach.

1	The crystal complexes represent biologically relevant structures and functions, especially the fatty acids tend to form randomly coiled conformations rather than discrete positions (driven by entropy) [2, 39].
2	Missing species data can be completed by computational means [34].
3	Differences in amino acids sequences between species explain the agonist-antagonism dualism.
4	Agonistic behavior of ligand other than LPS/LA may be due to contaminants in traces, i.e. false positive responses in low nonmolar ranges, e,g. Rhodobacter sphaeroides lipid A showed Chinese hamster agonism, but was tested as an murine antagonist and hence may be unreliable [10–13, 16] .
5	Agonist binding allows the heterodimerization of TLR4-MD-2-Lig complex.
6	Antagonist binding blocks the heterodimerization of TLR4-MD-2-Lig complex.
7	Docking and scoring show sufficiently responsiveness to reflect species differences in the sequences [14].
8	The torsion free energy can be estimated based on the 2D-connectivity graph of the ligand in a static way.
9	The side chain conformations of nonconserved residues can be repaired during protein homology modeling [35] and rearranged to reflect species differences upon docking [24, 25, 37].
10	The resolution of the crystal structures is sufficient allowing the positional elucidation of tiny electron densities corresponding to ligands’ alkyl chains in the hydrophobic patches of the MD-2 pockets, i.e. discarding artifacts through refinement software [2, 18, 39].
11	The acyl chains appear more deeply buried in the hydrophobic cavity of MD-2 in the case of antagonists like Lipid IVA and Eritoran, than lipid A/LPS [18, 19, 32].
12	The agonist position of lipid A/LPS with its protruding fatty acid FA1 is no artifact forced by crystal packing [18, 21, 39].
13	The reviewed mutation studies show no epiphenomena when associated with observed cell activity results [6].