Our Science & Technology
More signals. More information. Better context.
01 — Liquid Biopsy
A blood sample can contain molecular information.
Liquid biopsy provides access to DNA circulating in the bloodstream without requiring direct access to the tumour.
For CaniSense, we focus on cell-free DNA, analysing millions of circulating DNA fragments to search for patterns associated with cancer.
02 — Fragmentomics
It's not only about the DNA sequence. It's also about the fragments.
Cancer can alter the way DNA is released and fragmented in the bloodstream.
We therefore analyse characteristics such as:
Fragment length
Fragment-size distributions
Fragmentation patterns
DNA end motifs
Nucleosome-related patterns
Genomic positioning
These features can provide information about the biological origin and state of circulating DNA.
Fragmentomics is particularly interesting because some of these signals can be investigated even when the amount of tumour-derived DNA is very low.
03 — Genomic Signals
Cancer can change the genome.
Cancer cells can acquire genomic alterations as they develop.
One important signal is copy-number variation (CNV) — where sections of the genome are gained or lost.
By analysing genome-wide sequencing data, we can look for patterns of genomic instability that may contribute to the detection of tumour-derived DNA.
For CaniSense, CNV information can be combined with fragmentomic features rather than treated as a standalone signal.
04 — Multiple Biological Signals
Cancer is complex. One biomarker may not tell the whole story.
CaniSense is being developed around the idea of combining complementary molecular signals.
cfDNA
↓
Fragmentomics
↓
CNVs & genomic signals
↓
Additional molecular features
↓
AI / Machine Learning
↓
CaniSense
This gives the platform the ability to consider multiple dimensions of the underlying biology rather than depending on a single biomarker.
05 — AI & Machine Learning
Turning millions of DNA fragments into meaningful patterns.
Sequencing generates large amounts of molecular information.
Machine-learning approaches can analyse relationships across multiple features and identify patterns associated with different biological states.
Our research explores approaches including:
Random Forest
Gradient Boosting
Support Vector Machines
Deep Neural Networks
The objective is not simply to generate a score, but to develop models that can extract meaningful biological signals from complex cfDNA data.
06 — Explainable AI
Not just an answer. An explanation.
For a clinical decision-support system, knowing why a model produces a result is important.
Our approach therefore incorporates Explainable AI (XAI) to investigate which biological features contribute to a prediction and to help distinguish meaningful signals from technical artefacts.
The goal is to make AI outputs more transparent and useful for veterinarians.
This is also consistent with your existing positioning around trustworthy AI and the Fraunhofer HHI collaboration.
07 — From Science to CaniSense
Turning molecular signals into veterinary insight.
CaniSense brings together:
Liquid biopsy
+
cfDNA analysis
+
Fragmentomics
+
Genomic signals
+
Bioinformatics
+
AI & Explainable AI
to develop a blood-based approach to canine cancer detection.
Our vision
To make molecular information accessible to veterinarians and support earlier, better-informed investigation of cancer in dogs.