Why We Are Building Sybil

Biology has no shortage of data formats. What remains more difficult is expressing biological knowledge in a form that is at once biologically meaningful, readable by researchers and directly useful to computation.

This distinction matters. A nucleotide sequence can be represented perfectly without describing what a biological system is expected to do. A model can encode reactions without necessarily preserving the evidence from which a biological assertion arose. A database can connect thousands of biological entities while leaving important aspects of mechanism or experimental context implicit.

Noviota is developing Sybil as an experimental domain-specific language for exploring this problem.

The intention is not to replace the standards that already underpin computational and synthetic biology. It is to investigate whether a relatively concise, human-readable language can provide another useful layer through which biological systems can be described, inspected and eventually reasoned about computationally.

Starting from existing standards

Any serious attempt to improve biological representation should begin by acknowledging how much infrastructure already exists.

The Synthetic Biology Open Language (SBOL) provides a community-developed, machine-tractable representation for biological designs. SBOL3 substantially simplified the standard while extending its ability to represent biological designs across scales and through the design–build–test–learn process. Its ontology-backed model and use of Semantic Web technologies make interoperability a fundamental part of the standard rather than an afterthought. [1]

For computational models, the Systems Biology Markup Language (SBML) performs a different but equally important role. SBML Level 3 provides an extensible framework for exchanging mathematical models of biological systems, including reaction-based, constraint-based, logical and other forms of systems biology model. [2]

Provenance has its own mature standards. The W3C PROV-O ontology, for example, provides a general framework for representing entities, activities, agents and the provenance relationships between them. [3]

Sybil is being developed alongside this ecosystem, not in opposition to it.

Why another language?

The question behind Sybil is therefore not whether biology needs another file format.

It is whether there is value in a language through which a researcher can express a biological system in something closer to the way they reason about it, while retaining enough formal structure for software to validate, transform and interrogate what has been written.

Consider the difference between recording that two biological entities exist and making an explicit assertion about their relationship. For computational reasoning, we may eventually want to express not only that a pathogen protein interacts with a host protein, but the nature of that interaction, the biological context in which it occurs, the evidence supporting it and the consequences attributed to it.

As these descriptions become richer, readability matters. A representation intended to participate in scientific reasoning should ideally remain inspectable by the scientist.

Sybil is our attempt to explore that space.

A language, not a replacement standard

We currently think of Sybil as a domain-specific language for computational biology.

Its development is guided by several principles: biological entities and relationships should be explicit; representations should be composable; assumptions should be inspectable; validation should be possible before downstream computation; and interoperability with established standards should be preferred wherever practical.

SBOL3 is particularly important here. It already provides a sophisticated representation of biological design information, and software such as pySBOL3 demonstrates how that model can be manipulated programmatically. [4]

Sybil therefore does not need to reproduce SBOL.

Instead, one of the questions we are exploring is whether a more concise language can provide a useful authoring and reasoning layer, while allowing appropriate biological designs to be translated to or connected with established representations such as SBOL3.

The same principle applies elsewhere. Where a problem is properly represented as an executable systems-biology model, SBML already provides a mature exchange framework. Where provenance needs to be exchanged between systems, PROV-O provides an established vocabulary.

A useful language should know when to interoperate rather than reinvent.

From representation towards reasoning

The longer-term motivation for Sybil comes from Noviota’s wider research programme.

We are interested in a progression from:

Representation → Modelling → Reasoning → Design

Representation is deliberately first.

Before asking a computational system to reason about a biological mechanism, we need to know what has actually been represented. Which biological entities are involved? What relationships have been asserted? Under what conditions? On what evidence? Which statements describe observation and which describe inference or intended design?

These questions become particularly important in infectious disease and immunology, where biological behaviour depends heavily upon host, pathogen, cellular state, time and context.

Sybil is therefore an experiment in making some of that structure explicit.

We do not yet know what the final language should look like, nor do we assume that a single language can capture every useful representation of biology. Developing Sybil is itself part of the research: an opportunity to discover which biological concepts can be expressed cleanly, where existing standards already provide the better answer, and where the boundary between representation and reasoning begins to appear.

That is precisely why we are building it.

Related reading

Sybil is motivated by a broader argument that biological representation is not simply a data-formatting problem. The question of how much structure a computational model must capture also runs through our research note on what a genome model needs to represent in order to produce a working virus. Our colleagues at Molecular Precision examine structured biological design and reproducibility from the perspective of nanomedicine, where formal representation helps make increasingly complex interventions inspectable and reproducible.