Skip to content
antibacterial .ai

Lead optimization in drug discovery, when the target is a bacterium

Standard lead optimization pushes lipophilicity up. For Gram-negatives that is often exactly the wrong direction, and the series dies on the plate.

4 August 2026 9 min read Chemistry

Standard lead optimization has a well worn direction of travel: raise potency, tune lipophilicity upward for membrane permeation, manage the resulting solubility and metabolism problems. Applied to a Gram-negative antibacterial series, that playbook frequently walks the compounds off a cliff.

Why the usual direction is wrong here

Human cell membranes are lipid bilayers, so passive permeability correlates with lipophilicity. The Gram-negative outer membrane is not that. Its outer leaflet is lipopolysaccharide, densely packed and strongly anionic, and it is a poor environment for hydrophobic molecules. The realistic route in is through porins, water filled channels that favour small, polar, non bulky molecules.

So the property that helps a compound cross a human cell membrane is close to the property that stops it entering E. coli. Work on accumulation in E. coli has been summarised as the eNTRy rules: an ionisable Nitrogen, low three dimensional complexity, and limited Rotatable bond count tend to favour accumulation. Primary amines feature repeatedly in compounds that accumulate well.

Efflux is the second half of the same problem

Entry is necessary and not sufficient. RND efflux pumps span both membranes and export a wide range of substrates, so the concentration that matters is the balance between influx and efflux. Two consequences for a series:

  • A compound can look inactive on wild type and active on an efflux deficient strain. That difference is a measurement of the problem, and it is the single most informative experiment early in a Gram-negative series.
  • Chemistry that reduces pump recognition often moves in the same direction as chemistry that improves porin uptake: smaller, more polar, less flexible.

The pumps and their organisms are listed on the Gram-negative bacteria page.

A sequence that works better

  1. Establish whether the problem is affinity or exposure. Compare whole cell MIC against target inhibition, and wild type against an efflux deficient or permeabilised strain. Until this is known, potency optimisation is guessing.
  2. Fix exposure first. If accumulation is the problem, chase accumulation. Potency gained before the compound can get in is usually lost again.
  3. Keep a Gram-positive control in the panel. Activity on MRSA with nothing on E. coli is a clean uptake signature, since the Gram-positive organism has no outer membrane to cross.
  4. Watch resistance frequency, not only MIC. A potent compound that selects resistance at high frequency is a worse lead than a slightly weaker one that does not.
  5. Check the mechanism attribution on every loss. Efflux and permeability are chemistry problems. Target mutation usually is not. The antibiotic resistance page sets out which is which.

The properties worth tracking

PropertyUsual small molecule directionGram-negative antibacterial direction
LogDModerate to highLow, often below zero
Molecular weightUp to roughly 500Smaller is better, porin limits bite early
ChargeNeutral preferredAn ionisable amine is often helpful
RigidityFlexible linkers toleratedFewer rotatable bonds, flatter shapes
Efflux liabilityRarely measured earlyMeasured early, on purpose

Gram-positive series are a different exercise. Without the outer membrane, exposure is less often the limiting factor and conventional potency and property optimisation behaves more normally. This is one of the reasons a series should be screened against both.

Where triage fits

Most of the decisions above are cheap to make on paper and expensive to make on a plate. Screening the series before committing plate time gives you the ranking and, more usefully, the attribution: which compounds are losing to efflux, which to permeability, which to a target that has already moved. Then the plate confirms the top of a list that was built for a reason.

From the note to a read-out

Screen the compound this note made you think about. Three screens per session, no account, and the same matrix a paid run returns.

Run a screen first

What you are agreeing to

  • Your compounds stay yours
  • Never used for model training
  • Deletable on request
  • No card required to run a screen

Screen your own compound.
No card required.