GEPER develops computational systems for interpreting genomic variation, combining genomics, computational biology and artificial intelligence so that biological data becomes understandable, evidence-backed insight for the people who need it.
How the system works
The same five stages run through everything GEPER builds. Scroll to follow a single variant through them.
Stage
01Biological data
Genomic data arrives as sequence and as variants, positions where an individual differs from a reference, alongside population and molecular context.
02Computational representation
Each variant is placed in context: gene, transcript, protein region, population frequency, conservation. These become structured representations a model can reason over.
03AI and statistical models
Machine learning and statistical models estimate what a change might do to splicing, protein function or regulation. Each estimate is one line of evidence, not a verdict.
04Biological interpretation
Model estimates are weighed with curated clinical and population knowledge under established frameworks, with the reasoning recorded step by step.
05Research and clinical output
An interpretation a scientist can read, audit and challenge: what was concluded, what supported it, and what remains uncertain.
Research
How should sequences and variants be represented so that computation preserves their biological meaning?
A variant means different things depending on its gene, transcript, protein context and population background. This pillar studies how to encode that context consistently, and how to keep representations honest when parts of the picture are missing.
How can different sources of biological information be combined?
Sequence, structure, expression, conservation and population data each describe a different facet of a variant. This pillar explores how to combine such sources so that the whole is more informative than any single model.
How can genomic variation be connected to potential biological consequences?
Interpretation is where evidence becomes a conclusion. This pillar focuses on combining computational estimates with curated knowledge under established frameworks, and on making every step of that reasoning traceable.
Can computational models generate useful biological predictions and hypotheses?
Beyond interpreting known variation, GEPER is interested in whether models of biological sequence can propose hypotheses worth testing. This is a longer-term direction that depends on the foundations above.
Next
The Technology page walks through each of the five layers and the principles behind them.
