zartler chi fbdd_apr2013

20
“Fat, Drunk, and Stupid is No Way to Go through Life”: (Re)Thinking Fragment Libraries Edward R. Zartler, Ph.D Quantum Tessera Consulting And Chris Swain, Ph.D Cambridge MedChem Consulting

Upload: edward-zartler

Post on 09-Apr-2017

250 views

Category:

Science


0 download

TRANSCRIPT

Page 1: Zartler chi fbdd_apr2013

“Fat, Drunk, and Stupid is No Way to Go through Life”:

(Re)Thinking Fragment LibrariesEdward R. Zartler, Ph.D

Quantum Tessera ConsultingAnd

Chris Swain, Ph.DCambridge MedChem Consulting

Page 2: Zartler chi fbdd_apr2013

ENC 2004

2

Page 3: Zartler chi fbdd_apr2013

The “Rule” of 3

• MW< 300 (< 21 HAC)• H-BondDonor< 3 • H-BondAcceptor < 3• cLOGP< 3Also suggested:• # RotBonds< 3• PSA< 60 (hey, it’s a multiple of three!)

www.quantumtessera.com 3

Drug Discovery Today (2003) 8:876

Page 4: Zartler chi fbdd_apr2013

Beware multiple qualifiers.

“We carried out an analysis of a diverse set of fragment hits that were identified against a range of targets. The study indicated that such hits seem to obey, on average, a ‘Rule of Three’…”• One paragraph before, they mention only kinases and

proteases, so… is the Ro3 based upon only these target types? Seems so…

• If so, those who dogmatically use it are being stupid unenlightened, especially if they are screening proteases/kinases.

“The study indicated that such hits seem to obey, on average…”

www.quantumtessera.com 4

Page 5: Zartler chi fbdd_apr2013

Properties of Published Fragments

Slide from C. Swain 5

450 fragments against 67 targets40% contain ionizable group

Page 6: Zartler chi fbdd_apr2013

Because Animal House is full of Wisdom

6

MW/HAC cLogP

Do

FBHG

Page 7: Zartler chi fbdd_apr2013

Fat

• Arbitrary cutoffs are arbitrary and thus do not involve thinking

• According to a poll at the Practical Fragments blog, nobody uses fragments >20 HA

• Does these rules apply to Fsp3-enriched libraries?

• Rules should be applied when they are fully understood.

www.quantumtessera.com 7

Page 8: Zartler chi fbdd_apr2013

Drunk

• cLogP: the partitioning of an “un-ionized” compound into two immiscible solvents at equilibrium.Typically, water and alcohol (n-octanol).

– 1.4 logP units is added during optimization– Is solubility a better metric to track?

www.quantumtessera.com 8

J. Med. Chem., 2013, 56:2478

Page 9: Zartler chi fbdd_apr2013

State of the Art: Solubility

• Experimentally Measure, and first!

9E.R. Zartler et al. (2013) Drug Discovery World, Winter 2012/2013

Solu

ble

In- S

olub

le

Page 10: Zartler chi fbdd_apr2013

Stupid

• Dogma is static, does not take into account reality

• Every library has rules they apply:– RO3 being the main one.

• Everything should be done with a purposewww.quantumtessera.com 10

Page 11: Zartler chi fbdd_apr2013

What do medChemists think they need?

• Molecules with activity towards Target– Selectivity (Bonus!)– Selection of screen is key (Short Course)

• Novel Scaffolds– Is there such a thing as novelty in FBHG?

• Chemical space to work in– Every atom is sacred

• SAR– I like it early, helps to confirm screen hits– Many people will generate it later

Page 12: Zartler chi fbdd_apr2013

12

MedChem is All the Same

DKd DDG10x 1.37100x 2.731000x 4.2

DG=-RTlnKd

R=1.99 cal/KmolRT=0.6 kcal/mol at 300K

10x = 1.4kCal=1 Hydrogen BondIdeally, 0 atoms have to be addedRealistically, 3 atoms give10x (LEAN of 0.3)

The difference between 100mM and 10mM isone different atom, or one ideal atom, or three good ones.

It’s all Thermodynamics

1mM->1nM is 18 good atoms more (31 atoms at 1nM)1mM->1nM is 9 good atoms more (35 atoms at 1nM)

Beware the Sauron Atom!http://practicalfragments.blogspot.com/2012/11/atoms-are-like-apples.html

Page 13: Zartler chi fbdd_apr2013

Cascading Fragment Libraries

13

Binding Assay 1 (100 cpds)

Binding Assay 2 (1000 cpds)

Biochemical Assay (>10,000 Cpds)

Page 14: Zartler chi fbdd_apr2013

The Role of 19F

• 19F is exciting and a personal fetish– See Brad Jordan’s talk from FBLD last year!

• Where/when would you use it?• Aliphatic libraries because NMR-focused

libraries like to avoid a lot of aliphatics?• Target focused screening 19F-protein

www.quantumtessera.com 14

Page 15: Zartler chi fbdd_apr2013

15

N

N

S

2D Fragments

• Planar fragments seemed to be adored by certain classes of targets, e.g. kinases.

• Planar fragments explore “chemical space” efficiently, but what about “vector space”?

– Bond rotations and global motion float, interaction sites are still relatively limited.

• Are we like Khan, do we think two-dimensionally to a fault?

N N

S

Page 16: Zartler chi fbdd_apr2013

3D Fragments• Seem to be loved by certain target classes• Sample chemical space inefficiently• Are able to sample vector space to varying degrees.• 3D spiro compounds vs. rotatable bonds

• Rotatable bonds pay an entropy cost, but access more vector space. More rotation, more entropy, more vector space…

N

N

Page 17: Zartler chi fbdd_apr2013

17

3D fragments expectations

• If you have structural data (X-ray), 3D fragments should only be used as a tool.

• If you don’t have structural data, the fragments should be part of the final molecule, or you at least need to have a strategy for combining it with other fragment hits.

• 3D libraries are bigger (in general) than 2D fragments…so you start with fewer addable atoms.

• A wholy 3D fragment library would have to be FAR larger than an equivalent 2D library.

http://3dfrag.org/

Page 18: Zartler chi fbdd_apr2013

18

2D vs. 3D Fragments

• Is this an either/or proposition?– Shouldn’t be…like anything its compromise

• The ideal library would be:– Primarily 2D fragments (very planar) “Chemical Space”– Large enough to cover sufficient chemistry space– Built-in SAR to develop hypotheses.– Small portion of 3D fragments, but related to the 2D members

“Vector Space”

– Rapidly accessible follow up fragments for SAR

Page 19: Zartler chi fbdd_apr2013

www.quantumtessera.com19

“3D-arity” of Commercial Libraries

Page 20: Zartler chi fbdd_apr2013

Principal Moment of Inertia (PMI)Better 3D Metric?

20

IotaLead-Like:CCG

WIREs Comput Mol Sci 2012, 2: 868–885