The halo(gen) effect in para substituted phenyl rings - EuroCup2015

The Halo(gen) Effect in Para-Substituted Phenyl Rings

or

Understanding our Love Affair with para-Chlorophenyl: Scientific Rationale or Unsubstantiated Bias?

Dean G. Brown, Moriah M. Gagnon and Jonas Boström

Infection Innovative Medicines, AstraZeneca R&D Boston, Waltham, 02451.

CVMD Innovative Medicines, AstraZeneca, Mölndal SE-431 83, Sweden

I read a book – I write a blog-post

…and I get my own banner

1. Not even wrong - The Failure of Drug Likeness Rules

drug-likeness rules are controversial for many reasons..

2. Copy More! Copy Better!

Pharma to re-examine its aversion to copying, and 'follow-on' drugs are good for companies and patients

3. Searching for E.T. in Space and Drugs in the Cloud

The time for cloud computing has come and the difference it will make will be revolutionary

4. Luck, Molecular Modeling and the Sports Illustrated Curse

Drug design and Sports Illustrated connected via the statistical concept of 'regression to the mean’

5. A World Without (New) Drugs – Play It Before You Live It?

Gamification is trendy, examples of its employment have produced incredible results

6. Compulsive Gambling, Pregnancy Scores and Matched-Molecular Series

How Nassim Taleb's Black Swan relates to the topic of drug discovery and how it could help

7. Don’t Push That Single (Placebo) Target!

Is the prevailing one drug, one target ideal just an illusion?

8. The Halo(gen) Effect in para-Substituted Phenyl Rings

Decisions may well be based on personal preferences and intuitive biases.

This is not a book…

Last blog-post not from a book… but yet a very interesting observation by colleague

Dean Brown, AZ Boston

The latest blog-post – on The Halo(gen) Effect

The halo effect is a specific type of confirmation bias that makes us perceive someone (or something)

favorably because of one very positive quality in that person/thing.

“That Hollywood actress is beautiful she must also be clever/happy/[fill in the blank]“.

The Halo Effect – I needed an angle to football

Attributions based on data that we for some reason believe are reliable

…but can cloud our judgment and infer with decision-making

“David Beckham is looking good and few in the

world can kick the ball like him. He's likable

and extremely popular. It is easy to think that

he's all good.

But…he has recently been accused of triggering

a halo effect around unhealthy drinks by

endorsing Pepsi”

Ice-pick lobotomy anyone?

The Halo Effect – I needed an angle to life science

• An ice-pick lobotomy technique developed making unruly patients calmer

• Without all the facts this technique was used – it gave the expected results

• 20 000 people ‘treated’ before science corrected itself

• Common in science – we need to question more…?

Dr. Walter Freeman (1895 – 1972) drove a van (lobotomobile) around the US teaching his technique

The Halo(gen) Effect in Para-Substituted Phenyl Rings

Understanding our Love Affair with para-Chlorophenyl: Scientific Rationale or Unsubstantiated Bias?

AZ Med Chem Survey – 50 medicinal chemists

expected an even distribution but...

“a strong preference toward para substitution in phenyl rings was revealed”

Web-survey

Responses (<4h)

“Survey says”

“…that is where metabolism is most likely”

“…previous experience tell me (gut feeling) will most likely be the most potent one”

“…my experience tells me that this position is most likely to boost the potency”

“…ortho will have an effect on conformation as well as interactions with the Cl or R. Meta always feels a bit off”

“…the para is the easiest to make (in most cases) and would often give a boost in potency but also lipophilicity”

Para, because…..

What is true vs anecdotal?

How prolific is the para-Bias?

It is found across EVERY database we examined…

What about other 'R-groups'?

• Bias is found across most aromatic substitution patterns • Two exceptions shown for CF3 and pyridine.

Protein classification & frequency of aromatic substituents in PDB

Different bias according to target class?

Variations by target class for CF3 and pyridine, not p-ClPh.

The bias of m-CF3 Ph due to a high proportion of transferases (includes kinases). para-substituted pyridines often avoided to steer clear from CYP inhibition

Possible reasons for the bias?

• The Topliss work promoted para-substitutions? • DMPK (solubility, metabolism) effects? • Safety (hERG) property differences? • Ligand-binding effects (potency, ‘privileged’-fragments)? • Synthetic feasibility/accessibility? • Cost differences for chemical reagents? • Historically different design strategies (classic pharmacology vs. target-based design)? Personal preferences, subjectivity….or superstitions?

Does this really matter?

Why do we make certain decision/selections in drug design?

- Experience - Expediency - Conventional wisdom - SAR or models - …

Drug design is not an exact science. It is an aggregation of many parameters. Could we do better?

How do we better understand?

• Using data – many corporate collections are large enough for retrospective studies (e.g. reagents)

• What have we learned about reagents?

- Superior? - Over-hyped? - Which have proven problematic? - Over-sampled without good reason?

Beware of anilines, unsubstituted furans, PAINS...?

Synthetic accessibility may play a role

Synthetic accessibility or the perception of synthetic accessibility is the cause for the bias? References available where hindered amines react under typical amide bond forming conditions with good yields. the m-tBu are also underrepresented, not a steric influence... If synthesis difficult due to steric crowding --> an opportunity to less planar compounds and geometric diversity Beak, P. et al “Directed lithiations: The effect of varying directing group orientation on competitive efficiencies for a series of tertiary amide, secondary amide, and alkoxide directed ortho lithiations. J. Am. Chem. Soc. 1993, 115, 10628−10636

• In 2014 most common regioisomeric reagents are commercially available and cheap • In 1979, this was not the case for sulfonamides and boronic acids

• Little cost difference exists between the benzoic acids and aniline Cl-Ph regioisomers • para substituted sulfonyl chloride is 10-fold cheaper, but the price per gram ($2.55)

affordable even in the tightest of pharmaceutical budgets ....likely ruling out a price based reason.

Cost/Availability Play a Role?

Aldrich catalog on eBay in mint condition

How Influential was the Topliss paper?

Did the Topliss paper cause the bias, or reinforce current viewpoints?

“Since many systems are +p dependent, i.e. activity increases with increasing p values, the p-chloro analog is a good first choice, particularly since the ease of synthesis, relative to other substituted phenyl

compounds, is generally favorable” – Topliss 1972

Cumulative frequency over time (1960 and 2014) within the AZ corporate collection for -Cl, -F, -Me, and -MeO substituted aromatics

AZ Collection: Time vs. Registration

Significant bias for para substitution for Cl, F, Me, and MeO-Ph substituted Using Python scripts and OEChem

(spike 1995 with the rise of chemical and biological technologies )

Frequency of Cl, F, Me, and MeO substituted aromatics between 1960 and 2014

Time vs. Registration -- Normalized

Even more prominent before Topliss paper - reflects a very different design strategies? Drug projects relied on in vivo efficacy screens to drive SAR, now a plethora of in vitro screens. Unclear why the para bias persisted, especially given the availability of common reagents

Matched Pair Analysis: Is p-ClPh a Privileged Scaffold?

Little difference in potency (pIC50) across all assays, which may help to dispel the medchem myth “ortho substitution will most likely reduce activity because of the change in conformation” Other p-ClPh slightly more metabolically stable than meta o-ClPh slightly more soluble, better for hERG

Sampling of binding modes of p-ClPh within the PDB.

p-ClPh = “Privileged” Binding Mode?

Cl-Ph regioisomers are different with respect to shape and electrostatics There was not a dominant binding mode of p-ClPh. Many different binding modes…

Analysis of 50+ PDB structures across select regioisomers

The most dominant amino acids were the aliphatic amino acid... All regioisomers prone to be buried deep inside hydrophobic pockets Little difference between Cl-Ph regioisomers and 4-pyridine

All of the amino acids within 5 Å of were tabulated frequency of occurrence was calculated.

p-ClPh = “Privileged” Binding Mode?

Distribution of ClPh regioisomers in 10 HTS active lists

What is the Impact of the Bias?

Summary

• No single good reason why pCl-Ph was found in 3:1 ratios in all databases. • para-bias effect observed for most other R-groups (F, Me, OMe,...) • Bias even more pronounced before Topliss within AZ

historically distinct medicinal chemistry design strategies?

• Perceptions reinforced around synthetic difficulty of meta and ortho? Why? • The continued accumulation of para compounds in collections reinforces the

(erroneous) bias that they are 'unique' • Using skewed molecular databases can be risky if one is not aware of the uneven

distributions (e.g. HTS triaging – more para hits due to probability) • The present day bias is likely due to unjustified personal preferences • There are opportunities for screening collection differentiation, and improved use of statistics

http://pubs.acs.org/doi/pdf/10.1021/jm501894t

The Next Blog-Post?

http://www.pharma-iq.com/columnists/on-drug-design/

Read a book – write a blog-post

Any recommendation on books to read?

jonas.bostrom@astrazeneca.com