The Science Behind Our Approach
Why longevity variants, why mRNA, and why the LNP-CNS delivery challenge is tractable — our reasoning, stated plainly.
The Comparative Genomics Hypothesis for Longevity
The hallmarks of aging framework — described by López-Otín and colleagues in Cell (2013, updated 2023) — catalogs the cellular and molecular mechanisms underlying biological aging: genomic instability, telomere attrition, epigenetic alterations, loss of proteostasis, disabled macroautophagy, deregulated nutrient sensing, mitochondrial dysfunction, cellular senescence, stem cell exhaustion, altered intercellular communication, chronic inflammation, and dysbiosis. The framework tells us what goes wrong. It does not tell us which proteins, if upregulated, would slow that progression.
Evolution has already run longevity experiments at scale — across hundreds of millions of years and thousands of species. We are reading the results.
That is where comparative genomics becomes a discovery tool. Species like the naked mole rat (Heterocephalus glaber), bowhead whale (Balaena mysticetus), Greenland shark (Somniosus microcephalus), Brandt's bat (Myotis brandtii), and Aldabra giant tortoise have independently evolved exceptional longevity phenotypes across phylogenetically distant lineages. Genetic variants conserved across these species — variants showing signatures of positive selection in longevity-relevant contexts — represent evolutionary hypotheses about which protein functions matter for long life. The logic is analogous to how loss-of-function variants in PCSK9 identified in human populations with protected cardiovascular phenotypes became a drug target. We apply the same logic to longevity.
Comparative genomics for longevity is not a new idea. What caVos adds is systematic AI-guided prioritization — ranking variants by therapeutic tractability, then coupling that directly to mRNA sequence design and LNP delivery modeling. The discovery pipeline is computational; the validation requires experimental partners.
Why mRNA — and Why Not Gene Therapy or Small Molecules
Our targets require upregulation, not knockout. That constraint narrows the modality options considerably. Here is our reasoning — including the challenges we are not hiding.
- Transient expression — controllable dosing
- No genomic integration risk
- Programmable sequence per target
- Rapid design iteration in silico
- Established LNP delivery precedent (mRNA vaccines)
- Challenge: stability, delivery to CNS, immune activation
- Durable expression from single dose
- Genomic integration risk (insertional mutagenesis)
- Irreversible — correction requires additional intervention
- Immune responses to viral vectors
- Less suited to titrated upregulation
- Oral availability in many cases
- Cannot easily upregulate protein expression
- Target must have a druggable binding pocket
- Off-target selectivity challenge
- Often inhibits rather than upregulates
The Blood-Brain Barrier Problem
The blood-brain barrier (BBB) is a selectively permeable boundary maintained by tight junctions between brain endothelial cells, efflux pumps (P-glycoprotein, BCRP), and restricted transcytosis pathways. CNS drug delivery is genuinely one of the most difficult problems in medicine. Most systemically administered macromolecules — including LNPs — do not cross at therapeutically meaningful concentrations.
For lipid nanoparticles delivering mRNA, key formulation parameters that influence BBB penetration include: ionizable lipid pKa (affects endosomal escape), lipid molar composition (influences particle surface properties), PEGylation degree (affects circulation time and uptake), and particle size (influences biodistribution). Published work in the CNS-LNP space has begun to characterize which formulation windows correlate with brain uptake in animal models — establishing that the search space is bounded and systematic exploration is tractable.
caVos approaches this computationally: treating the formulation parameter space as a search problem, using predictive models to rank configurations before committing to wet-lab synthesis and testing. Our CVX-003 program is the computational and early experimental foundation for this approach. We don't claim a solved delivery problem — we claim a more systematic way to search for one.
Computational-First. Honest About What That Means.
“We are a computational-first lab. We do not run animal experiments. We design, model, and partner for validation.”
This is not a limitation we apologize for — it is a deliberate position. Computational biology has proven productive for hypothesis generation and candidate prioritization at a fraction of the cost and lead time of wet-lab-first approaches. The corollary is honest: our candidates carry no experimental biology yet. Computational predictions are starting points, not results.
When we say "discovery stage," we mean a candidate sequence has been designed and characterized in silico. No cell data. No animal data. No clinical data. We build computationally, then seek experimental partners who have the infrastructure to test what we've designed. Scientific credibility built on accurate stage representation is the only durable foundation for the pharma partnerships we are working toward.