In preclinical pharmacology, a biomarker is only as useful as its ability to inform a go/no-go decision. The question is never simply “does this compound change a number?” — it is “does this number tell us something about mechanism that will hold in the clinic?”

That distinction drives how we build our preclinical biomarker strategy at Ulysses Neuroscience.

In our IFN-α rat model of depression — one of the most pharmacologically validated inflammation-driven models available, replicated across more than 25 independent studies in our laboratory — we characterise the full molecular signature of the disease state and its reversal. That means inflammatory markers, BDNF, synaptic proteins, and α-tubulin post-translational modifications measured in plasma, prefrontal cortex, and hippocampus at study endpoint, alongside a behavioural profile that now extends well beyond the Forced Swim Test.  We have now validated this approach in two other models of depression, the Wistar Kyoto rats (naturalistic model) and the mouse chronic social defeat (psychosocial stress model). The in-vivo phase of the latter is performed by our partner Transpharmation.

Moreover we run naturalistic behavioural measures in both rats and mice — nest building, burrowing, marble burying, the splash test — in parallel with classical assays, because these readouts capture motivation, self-care, and species-typical functioning in ways that map more directly onto patient-relevant outcomes. Our team as also started to validate the measure and analysis of ultrasonic vocalizations (USVs) in mice under conditions of stress or when exposed to pleasurable stimuli.

This multimodal approach — molecular, synaptic, and behavioural — builds a compound fingerprint that travels across the translational pipeline rather than terminating at a single endpoint.

Electrophysiology adds a further dimension that we are actively integrating. Our EEG platforms — tethered, wireless, and video-synchronised — are currently validated in our Developmental and Epileptic Encephalopathies (DEEs) models, where Cdkl5-Het mice show elevated spectral power, particularly in theta, beta, and gamma bands, relative to wild-type animals. Implementation into the IFN-α rat model is underway, with first data expected in Q3 of 2026. The goal is the same across all models: a preclinical biomarker signature that does not merely describe an animal’s pharmacological response, but predicts what we expect to measure in human samples and, ultimately, clinical endpoints.

The challenge in translational medicine is making preclinical and clinical teams speak the same biological language. Preclinical and clinical programmes are often measuring biology at different levels of resolution, in different matrices, with different technologies — and then asking whether their findings align. The answer is usually no, and the result is clinical programmes that fail not because the biology was wrong, but because the translational bridge was never properly built.

Acetylated α-tubulin — a post-translational modification of the microtubule cytoskeleton reflecting neuronal plasticity and cytoskeletal integrity — is one marker where we have built that bridge systematically. In our IFN-α rat model of depression, acetylated α-tubulin levels are significantly increased in plasma and prefrontal cortex relative to saline controls — a disruption normalised by both ketamine and psilocybin. Importantly, our in-house data show that patients affected by major depressive disorder have increased acetylated α-tubulin in plasma and post-mortem orbitofrontal cortex, compared to healthy controls. In iPSC-derived cortical neurons, ketamine treatment produces a dose-dependent reduction in acetylated α-tubulin at 3 µM and 10 µM — consistent with a rapid cytoskeletal remodelling response. And in plasma from 49 CDKL5 Deficiency Disorder patients compared to 45 healthy controls, we observe a highly significant elevation in acetylated α-tubulin, reflecting the cytoskeletal dysregulation arising from CDKL5 loss of function that we also observed in similar fashion in the brain and plasma of Cdkl5 mouse mutants.

The magnitude of changes differs across these contexts and it is accompanied by differential alterations in inflammatory markers, neurotrophins, NfL and neurodegeneration related markers. Thus, the biomarker answers doesn’t rely in one single biomarker, but rather in the characteristic signature obtained by analyzing different biomarkers in the same sample. What is consistent is the signal itself: a quantifiable, cross-species molecular readout that is detectable in plasma, responsive to pharmacological intervention, and directly comparable between rodent models and human cohorts. That is the definition of a translational biomarker, and it is the standard we apply across our entire biomarker programme.

The technical infrastructure that makes this possible lives entirely within our Cherrywood laboratory: the Luminex 200 (flow cytometry) and the MSD QuickPlex SQ 120MM (electrochemiluminescence) for multiplex inflammatory markers, BDNF, NfL, neurodegeneration panels and metabolic markers. Li-Cor Odyssey CLx for infrared Western blot imaging of synaptosome synaptic markers and cytoskeletal microtubule proteins. All major platforms under one roof, on the same samples, with the same protocols — ensuring the consistency that multi-site programmes cannot replicate.

The conversation around biomarkers in neuroscience R&D is gaining momentum across the industry, and rightly so. We have been building this infrastructure quietly, study by study, data after data, validation after validation since Ulysses Neuroscience was founded. The data we are now generating — cross-species, cross-indication, across preclinical and clinical matrices — represents a level of translational integration that most CROs cannot offer, because it requires both divisions to operate as one.

If you are designing a CNS programme and want to understand how a rigorous biomarker strategy could de-risk your path to clinic, we would be glad to talk.