Why digital biomarkers

From a single endpoint to a continuous signal

A classic behavioural test gives one number, one morning, with a handler in the room. The DVC instead records activity 24/7, for weeks, undisturbed — and that continuous stream can be mined for many validated biomarkers, several of them detectable days to weeks before classical signs appear. The rest of this deck is a tour of those biomarkers.

Iannello 2019 · Pernold 2023 [1, 2]

How the biomarkers are built

One activity stream, several biomarker families

Iannello 2019 · Pernold 2023 [1, 2]

Biomarker · Activity & rest architecture

The daily activity budget

The foundation readout is the structure of the day: how far a mouse travels, and how that time splits into active bouts and rest. Deviations from a strain's normal budget — less distance, daytime activity, broken rest — are the baseline that disease- and drug-response biomarkers build on. Combined with body weight, that deviation already grades disease severity in models such as colitis, and it lives mostly in the dark phase that daytime testing misses [9].

Pernold et al., PLoS One 2023 · Zentrich et al., Front Neurosci 2021 [2, 9]

Biomarker · Rest Disturbance Index (RDI)

Fragmented rest, read without EEG

RDI captures how broken-up rest is — its regularity, not its quantity — derived from the entropy of the minute-by-minute activity trace. It is the closest thing the field has to a transdiagnostic digital biomarker: the same signal flags ALS, myotonic dystrophy, narcolepsy and ageing, often before motor symptoms, with no electrodes on the animal.

Golini et al., Front Neurosci 2020 · Piilgaard et al., Sleep 2023 [3, 10]

Biomarker · Circadian disruption

Body-clock biomarkers at rack scale

Treating the activity trace as a rhythm yields the classical chronobiology readouts — period, phase, amplitude, day/night ratio — recovered reliably enough to phenotype clock-mutant mice. Per-cage programmable LEDs make each cage its own light-controlled chamber, so jet-lag, free-run and shift-work protocols run without a dedicated darkroom.

Tir et al., Sci Rep 2025 [4]

Biomarker · Spatial signature

Frontality & wall activity

Because the floor is a grid, activity also has a place. Time spent against the walls and corners versus the open front ("frontality") is an anxiety-like / thigmotaxis biomarker — and the pattern is a strain fingerprint, distinct across C57BL/6, BALB/c and CD1. It works in group-housed cages and can be mined from archived recordings, with no new animals.

Fuochi, Rigamonti et al., Sci Rep 2023 [5]

Biomarker · Sleep proxy & narcolepsy

A sleep/wake proxy from the nest

A nest-identification step lets the cage separate inactivity-in-nest (a sleep proxy) from active wake — and it correlates with EEG/EMG-scored sleep. In narcolepsy type 1 models a strikingly simple marker emerges: an inability to sustain activity for more than ~40 minutes, detectable within the first weeks of disease onset.

Piilgaard et al., Sleep 2023 [10]

Biomarker · Metabolic & welfare

The bedding is a biomarker too

Brachs et al., Lab Anim (NY) 2025 · Collins et al., JAALAS 2025 [7, 8, 12]

Biomarker · Voluntary effort

When bulk activity isn't enough

Some phenotypes hide in motivation, not gross movement. Adding an in-cage running wheel turns voluntary running distance into a biomarker: in a cancer-induced bone-pain model, wheel running fell with the disease and tracked limb-use scores while general home-cage activity did not — pinpointing exactly when the wheel earns its place.

Hopkins et al., In Vivo 2025 [13]

An honest caveat

Group cage → individual?

Most of these biomarkers are read at the cage level. To get a per-animal value from a group cage, the usual shortcut is to divide by the number of mice — but a cage aggregate isn't cleanly one animal's: mice egg each other on, or huddle together to rest. It's a defensible average, validated against beam-frame systems, but worth stating as an assumption rather than a true individual measurement.

Sun et al., Front Neurosci 2024 · caveat: Neuronautix [6]

Making biomarkers reusable

A biomarker is only as good as its labels

One practical note for anyone building on these. The raw stream is exportable, but a biomarker value is meaningless without its context — strain, sex, age, housing, cage-change and light schedule, intervention timestamps. Capturing that metadata is what lets a biomarker travel between studies and feed virtual control groups. It is the quiet prerequisite behind everything above.

Big-data & standards review: Fuochi et al., Front Big Data 2024 [11]

The takeaway

The DVC is a biomarker discovery platform

That last step is where Neuronautix helps.

Where this comes from — the published studies

Sources