Conformational stability
Designing sequences that hold their intended fold rather than drifting between states.
Luminome
Peptides,
reimagined.
We design molecules for the language of skin.
Luminome combines computational intelligence and peptide biology to create a new generation of skincare ingredients.
The premise
Skin communicates
in molecules.
We design new ones.
Skin is a signalling environment. Cells coordinate renewal, barrier maintenance and the organisation of the extracellular matrix through short protein fragments that bind their partners with real specificity. Peptides are that vocabulary.
Cosmetic science has borrowed a handful of them — the ones that happened to be found first, in tissue, decades ago. They work because evolution shaped them for something adjacent to what we want.
The rest of the vocabulary has never been written down. That is the part we are interested in.
The platform
A design loop, not a screening campaign.
- 01
Understand
Models trained on peptide sequence, structure and measured behaviour learn how a change of one residue propagates into folding, binding and stability.
- 02
Generate
Rather than screening what already exists, we propose sequences that satisfy several objectives at once — shape, charge, solubility, robustness.
- 03
Validate
Candidates are synthesised and measured. Every result, positive or negative, returns to the model as training signal.
Signature method
We don't search for
ingredients. We design them.
A peptide is a sentence written in twenty letters. Its meaning — how it folds, what it binds, how long it survives — is set by the order of those letters. Reading that relationship in reverse is what our models do.
- 01 Sequence Twenty residues. A string of symbols.
- 02 Properties Grouped by charge, polarity and hydrophobicity.
- 03 Structure Folded into the conformation those properties imply.
- Hydrophobic
- Polar
- Charged
- Conformational
Featured design record
LM-0417
A helical fourteen-residue candidate, designed against a matrix-organisation objective and optimised for stability in an aqueous cosmetic base.
- Class
- α-helical peptide
- Length
- 14 residues
- Sequence
- K L A E T V R S L I D K A F
- Design objective
- Selective binding, retained fold, protease resistance
- Intended cosmetic function
- Topical ingredient supporting the look of skin firmness
- Status
- In evaluation
Illustrative of the design records our platform produces. The sequence shown is a representative example, not a disclosed Luminome candidate.
Design space
Almost none of it has been looked at.
A peptide of twenty residues has 2020 possible sequences — around 100 septillion. Classical discovery walks into that space by hand and samples a few hundred at a time, guided by what has already been found in tissue.
Generative models let us treat the space as something to be navigated rather than stumbled through: propose broadly, rank against every objective at once, and take only the few candidates worth making.
From sequence to skin
A designed molecule still has to survive contact with reality.
-
In silico design
Objectives set, candidates generated and ranked.
-
Synthesis
Solid-phase synthesis and purification of the selected set.
-
Biochemical testing
Binding, stability and solubility measured directly.
-
Skin models
Reconstructed human epidermis and ex vivo tissue.
-
Formulation
Compatibility, stability and sensory work in finished bases.
Evidence
One standard,
applied to everything.
Cosmetic science has a credibility problem, and it is largely self-inflicted. We would rather publish less and mean it.
Vehicle controls on every assay
A treatment arm means nothing without the matched base it was delivered in.
Reported n and dispersion
Sample size, replicates and spread are published with the central estimate, never after it.
Endpoints fixed in advance
The measurement is chosen before the data exists, so a result cannot be selected into significance.
Independent replication
A finding becomes a claim only once it has been reproduced outside the group that found it.
Methods published with results
Protocols, concentrations and models are disclosed so the work can be checked.
Study data will be published on this page as work completes, with methods attached. We have not put placeholder charts here in the meantime.
Capabilities
What the system optimises against.
Proteolytic stability
Backbones and residue choices selected to resist the proteases a topical peptide meets.
Aqueous solubility
Charge and hydrophobic patterning tuned so a candidate stays in solution at formulation pH.
Target selectivity
Shaping the binding interface toward one partner and away from close structural relatives.
Synthetic accessibility
Constraining designs to sequences that can actually be made at scale and at cost.
Formulation compatibility
Screening against the surfactants, emulsifiers and preservatives real products contain.
Mission
A new language for skin.
The most interesting molecules for skin have probably not been discovered yet, because discovery was never the constraint — imagination was, and after that, the sheer size of the search. Both of those have changed.
Luminome exists to design the ingredients that come next, and to hold them to a standard worth trusting.
Let's design
something new.
We work with formulators, brands and research groups who want ingredients that did not exist before.
hello@luminome.bio