MakeTNT-toolkit User Documentation	To MakeTNT-manual
MakeTNT Evaluation

MakeTNT Test Runs

Evaluation : Generation of TNT-Restraints by MakeTNT

No. Comp.	Restraint Type	Restraint Source	Successful Conversion	Conversion Failed	Reason for Failure	Remarks
49967	XPLOR_CORINA	NCI open source database (249071 3D structures) / Corina ^[1]	49967	-	-	-
2426	REFMAC	Refmac standard library ^[2]	2426	-	-	-
3289	XPLOR_XPLO2D	NCI open source database / xplo2d ^[3]	3288	1	input: 2 molecules	^[4]
1000	REFMAC (written by Corina)	NCI open source database / Corina ^[5]	1000	-	-	-

[1]	From the first 50000 compounds of the NCI open source database, 49967 structures with 3D coordinates have been selected. 10 of the structures have kindly been provided by Dr. Armin Ruf, Roche, Basel. XPLOR-type .top and .par files have been obtained from Corina.
[2]	All 2457 cif monomer descriptions (ccp4-4.2.2) have been run through Libcheck, Vers 4.0.1. 2426 lib-files could be obtained. For 31 files Libcheck failed.
[3]	From the first 20000 compounds of the NCI open source database, 3289 compounds have been selected if they had 3D coordinates and fulfilled at least one of the following criteria: > 1 4-atom ring, > 2 5-atom rings, > 2 6-atom rings, > 4 (4-6)-atom rings.
[4]	Compounds including metal atoms can cause trouble, as metal coordination is treated as a covalent bond by xplo2d, which increases the number of ring systems to be processed by MakeTNT. In order to cope with this problem, MakeTNT has been configured to process generous numbers of ring systems. In case of problems, please make sure, that the metal coordination listed in the *.top files generated by xplo2d is correct. If it still does not work, let us know.
[5]	The test was run on Refmac-dictionaries generated by Corina for the first 1000 coordinates of the NCI-dataset.

Evaluation : Small Molecule Refinements Using Corina-Derived Restraints Converted to TNT-Format by MakeTNT

Each cell in the table below represents one optimisation run, using the TNT-GEOM module without X-ray terms and a fast optimization routine kindly provided by Dr. Oliver Smart, Global Phasing. Corina generated coordinates have been randomized to a RMS deviations of 0.25, 0.5 and 1.0 Ångstrøms and optimized with TNT-dictionaries, sizes s, m and l. Different coordinates have been generated for each run. Each resulting structure has been manually inspected for integrity. Optimizations producing apparent errors in the geometry have been marked in red. The results indicate that:

Corina-generated XPLOR-restraints converted by MakeTNT reproduce correct structures, even after severe distortion.

Clustering of planar restraints (restraint size s(default), m) improves optimization.

-size "s"(default) produces the best results, especially if the starting model is very distorted.

-size "l" should not be used for restraints derived from XPLOR-type formats.

compound
number of atoms structure optimisation: jiggle[Å]/Size of MakeTNT-Restraint file
no. refinement cycles/successful?

0.25/s 0.25/m 0.25/l 0.5/s 0.5/m 0.5/l 1.0/s 1.0/m 1.0/l

NCI124652/28 126
yes^[a] 57
yes^[a] 74
yes^[a] 49
yes^[a] 45
yes^[a] 167
yes^[a] 73
yes^[a] 108
yes^[a] 55
no

NIST04051410392D/10 16
yes 36
yes 17
yes 17
yes 81
yes 14
yes 36
yes 32
yes 29
yes

NCI16498/25 13
yes 16
yes 184
yes 17
yes 25
yes 86
yes 60
yes 88
yes 114
yes

NCI367798/48 116
yes 231
yes 494
yes 284
yes 219
yes >2001
no 98
yes 416
no 415
no

NCI376722/29 26
yes^[a] 21
yes^[a] 144
yes^[a] 18
yes^[a] 26
yes^[a] 62
yes^[a] 158
yes^[a] 205
yes^[a] 415
no

NCI46/20 46
yes 21
yes 21
yes 59
yes 117
yes 215
yes 72
yes 136
yes 243
yes

NCI56940/28 21
yes 134
yes 173
yes 21
yes 74
yes 173
yes 73
yes 258
yes 115
yes

NCI628503/58 147
yes 127
yes 140
yes 190
yes 151
yes 396
no 238
yes 223
no 624
no

NCI631303/29 21
yes 46
yes 88
yes 199
yes 25
yes 189
no 35
yes 131
yes 303
no

NCI640266/26 61
yes 72
yes 157
yes 31
yes 75
yes 125
yes 112
yes 127
yes 266
yes

compound number of atoms	structure	optimisation: jiggle[Å]/Size of MakeTNT-Restraint file no. refinement cycles/successful?
0.25/s	0.25/m	0.25/l	0.5/s	0.5/m	0.5/l	1.0/s	1.0/m	1.0/l
NCI124652/28		126 yes^[a]	57 yes^[a]	74 yes^[a]	49 yes^[a]	45 yes^[a]	167 yes^[a]	73 yes^[a]	108 yes^[a]	55 no
NIST04051410392D/10		16 yes	36 yes	17 yes	17 yes	81 yes	14 yes	36 yes	32 yes	29 yes
NCI16498/25		13 yes	16 yes	184 yes	17 yes	25 yes	86 yes	60 yes	88 yes	114 yes
NCI367798/48		116 yes	231 yes	494 yes	284 yes	219 yes	>2001 no	98 yes	416 no	415 no
NCI376722/29		26 yes^[a]	21 yes^[a]	144 yes^[a]	18 yes^[a]	26 yes^[a]	62 yes^[a]	158 yes^[a]	205 yes^[a]	415 no
NCI46/20		46 yes	21 yes	21 yes	59 yes	117 yes	215 yes	72 yes	136 yes	243 yes
NCI56940/28		21 yes	134 yes	173 yes	21 yes	74 yes	173 yes	73 yes	258 yes	115 yes
NCI628503/58		147 yes	127 yes	140 yes	190 yes	151 yes	396 no	238 yes	223 no	624 no
NCI631303/29		21 yes	46 yes	88 yes	199 yes	25 yes	189 no	35 yes	131 yes	303 no
NCI640266/26		61 yes	72 yes	157 yes	31 yes	75 yes	125 yes	112 yes	127 yes	266 yes

[a]: structure of good quality, except for positioning of metal ions(e.g. Cu, Co), where currently no parameters for non-bonded interactions are available in Buster/TNT

Evaluation : Small Molecule Refinements Using Refmac-Derived Restraints Converted to TNT-Format by MakeTNT

Each cell in the table below represents one optimisation run, using the TNT-GEOM module as described in the previous section. 10 structures have been chosen from the Refmac (ccp4-5.0.2) standard library. Starting coordinates and library description files have been obtained by Libcheck (Vers 4.1.3). The coordinates have been randomized to RMS deviations of 0.25, 0.5 and 1.0 Ångstrøms and optimized with TNT-dictionaries, sizes s and l. Different coordinates have been generated for each run. Each resulting structure has been manually inspected for integrity. Optimizations producing apparent errors in the geometry have been marked in red. We observed that:

Refmac restraints not as good Corina restraints.

The two dictionary sizes l,s perform similarly well.

For the steroid thrombin-inhibitor compound (GR3), the problems are due to a the way planarity restraints are defined for the double bond in the six-membered ring. In this case, the problem can be spotted by manual inspection of the 3D coordinates generated by Refmac.

compound number of atoms	structure	0.25/s	0.25/l	0.5/s	0.5/l	1.0/s	1.0/l
compound number of atoms	structure	optimisation: jiggle [Å] / Size of MakeTNT-Restraint file no. refinement cycles / successful?
AQS 28		16 yes	13 yes	30 yes	40 yes	30 yes	46 yes
BRL 25		28 yes	55 yes	36 yes	34 yes	45 yes	36 yes
DPS 36		55 yes	93 yes	24 yes	106 yes	100 yes	291 yes
EST 20		36 yes	24 yes	53 yes	29 yes	90 no	86 yes
GR3 39		73 no	96 no	73 no	294 no	169 no	529 no
MTB 22		no atom labels in Refmac input (e.g.: 'C5'') currently not recognized by TNT
NEV 20		80 yes	34 yes	102 yes	71 yes	106 yes	29 yes
TMF 33		40 yes	105 yes	72 no	56 yes	99 yes	136 no
TXL 58		225 yes	257 yes	206 yes	155 no	395 yes	273 no
ZYA 28		56 yes	22 yes	95 yes	44 yes	61 yes	71 yes

Maria Brandl,<buster-develop@GlobalPhasing.com>

Last modification: 13.12.05