De Novo Design Software — Scoring-Strategy Catalog

P0 reference — verbatim Table 1.3 from 2014-schneider-de-novo-molecular-design-book §1.6 (Schneider & Baringhaus, pp. ~33–35). Chronological catalog of named de novo drug-design software (1989–2013) with the implemented compound-scoring strategy (receptor-based, ligand-based, or both). Useful for citing prior art when an STRC scoring script claims to follow a specific scoring class.

The two element-occurrence statistics from the same chapter (Table 1.2 — H 46.2%, C 40.0%, O 6.2%, N 5.6%, F 0.8%, S 0.6%, Cl 0.4%, P 0.1%, other 0.1% — across 12,647 drugs in COBRA v12.6) and the GDB chemical-universe estimate (166 × 10⁹ molecules ≤ 17 heavy atoms) bracket the scope of the search problem.

Verbatim Table 1.3

De novo design method	Year of publication	Receptor-based	Ligand-based
HSITE/2D skeletons	1989	X
3D skeletons	1990	X
Builder v1	1992	X
LUDI	1992	X
NEWLEAD	1993	X
SPLICE	1993	X
GroupBuild	1993	X
CONCEPTS	1993	X
SPROUT	1993	X
MCSS and HOOK	1994	X
GrowMol	1994	X
Chemical Genesis	1995	X	X
PRO_LIGAND	1995	X	X
SMoG	1996	X
CONCERTS	1996	X
RASSE	1996	X
PRO_SELECT	1997	X
Skelgen	1997	X	X
Nachbar	1998		X
Globus	1999		X
DycoBlock	1999	X
LEA	2000		X
LigBuilder	2000	X
TOPAS	2000		X
F-DycoBlock	2001	X
ADAPT	2001	X
Pellegrini and Field	2003	X	X
SYNOPSIS	2003	X
CoG	2004		X
BREED	2004		X
Nikitin	2005	X
LEA3D	2005	X
Flux	2006		X
FlexNovo	2006	X
Feher	2008		X
GANDI	2008	X	X
COLIBREE	2008		X
SQUIRRELnovo	2009		X
Hecht and Fogel	2009	X	X
FOG	2009		X
MED-hybridize	2009	X
MEGA	2009	X	X
Fragment-shuffling	2009	X	X
AutoGrow	2009	X
NovoFLAP	2010		X
PhDD	2010		X
GARLig	2010	X
DOGS	2010		X
White and Wilson	2010		X
Qsearch	2011		X
EvoMD	2011		X
Contour	2012	X
MOEA	2013		X
Ulrich	2013	X

— Verbatim from Schneider 2014, Table 1.3, Ch. 1 (adapted in turn from Ref. [162] of Ch. 1).

Scoring-class equations (referenced from Schneider 2014 §1.6.1)

Force-field-based receptor scoring (Eq. 1.10):

E = Σ_{i∈ligand, j∈receptor} [ A_ij/r_ij^12 − B_ij/r_ij^6 + (q_i q_j)/(ε r_ij) ]

Empirical scoring function (Eq. 1.11):

ΔG = ΔG_0 + Σ_i [ΔG_i · count_i · penalty_i]

Knowledge-based scoring function (Eq. 1.12, Boltzmann-inverted atom-pair statistics):

E(i,j) = −k_B T ln [ p_ij^observed(r) / p_ij^expected(r) ]

How to use in STRC

• h01 phase 4 docking pipeline: Vina is a force-field-based scorer (Eq. 1.10 family); cite Schneider 2014 §1.6.1 in phase4b.py docstring. AutoDock-class scoring is also force-field-class.
• h01 phase 4f MM-GBSA: force-field family with continuum solvent correction. Cite Schneider 2014 §1.6.1 + §16.2.5.2 (Westermaier & Hubbard) for the empirical-correction layer.
• Knowledge-based scoring lineage: SMoG (1996) is the first knowledge-based de novo scorer. If h01 ever moves to knowledge-based scoring (e.g., DSX-CSD), this lineage can be cited.
• DOGS (Schneider’s lab, 2010): ligand-based, fragment-grow with reaction-driven assembly. Same lineage as h01 v4 scaffold-hop attempts. Provides a literature precedent.
• LigBuilder (2000): receptor-based fragment-grow, used in Schneider 2014 §1.7 to discover EYA2 inhibitor 8 (IC50 = 6 µM) and BRAF inhibitor 9 (IC50 = 0.4 µM). Cited as a viable in-pocket fragment-grow method — relevant if h01 v4 fragment-grow is implemented from scratch.

Connections

• [source] 2014-schneider-de-novo-molecular-design-book
• [applies] index
• [see-also] Recipe — Receptor-Based Scoring Function Selection
• [see-also] Recipe — Fragment Optimization Linking Merging Growing
• [see-also] STRC Computational Scripts Inventory