Research
Advancing computational science through rigorous research.
Spacefield conducts research in computational molecular science, mathematical modelling and scientific computing to develop transparent computational methods, reproducible software and scientifically grounded technologies that support molecular discovery and experimental research.
Our research connects theoretical foundations, computational methods, software implementation, validation and publication so that scientific claims remain linked to reproducible evidence.
Evidence
Traceable computational research
Theory
Code
Data
Models
Benchmarks
Reports
Research philosophy
Research should produce knowledge that can be examined, reproduced and improved.
Spacefield conducts computational research that combines mathematical modelling, scientific computing and computational molecular methods to develop transparent, reproducible and scientifically traceable technologies.
Research begins with a scientific question and progresses through mathematical development, computational representation, software implementation and controlled evaluation. Where empirical data are used, models are benchmarked and interpreted within clearly defined scope and limitations.
Computational methods are designed to support scientific investigation rather than replace experimental evidence. The value of the research lies not only in predictive performance, but also in whether methods, assumptions, artefacts and results can be examined, reproduced and improved.
Research method
A disciplined path from concept to scientific release.
Each programme may differ in scientific detail, but the underlying research process follows a consistent sequence of definition, implementation, evaluation and release.
Define the question
Specify the scientific endpoint, intended use, assumptions and limits of the investigation.
Formulate the representation
Develop mathematical, structural or mechanism-informed quantities relevant to the problem.
Implement and test
Translate the representation into software with unit tests, invariance checks and failure handling.
Benchmark
Compare against appropriate baselines using predefined data partitions and evaluation metrics.
Validate
Assess generalisation, uncertainty, reproducibility and known limitations under controlled conditions.
Freeze and report
Preserve code, model identity, manifests, hashes, documentation and scientific conclusions.
Research programmes
A connected research portfolio from theory to applied molecular technologies.
Spacefield's research programme connects fundamental mathematical work, computational molecular methods, software implementation and applied prediction technologies. Each programme contributes to a broader scientific architecture rather than operating as an isolated research activity.
01
Foundational research
Fundamental Research
Theoretical and mathematical work that develops the formal foundations used across Spacefield's computational research.
- Spacefield Transformation Theory
- Mathematical modelling
- Operator-based representations
- Computational foundations
Public papers will distinguish established derivations, modelling assumptions, conjectures and proposed interpretations.
02
Computational research
Computational Molecular Research
Research that translates mathematical and scientific concepts into molecular representations, descriptors and reproducible computational methods.
- SFT Molecular Foundations
- SFT-Core
- Descriptor engines
- Molecular representations
Descriptor usefulness is assessed empirically; theoretical motivation alone does not establish predictive value.
03
Platform research
Applied Molecular Technologies
Endpoint-specific computational methods developed as research software and integrated into the SFT Molecular Platform.
- SFT-pKa
- SFT-logP
- SFT-Solubility
- SFT-BBB
- SFT-ADMET
- SFT-Binding
Scientific position: Computational research and prediction support scientific investigation, candidate prioritisation and hypothesis generation. They do not replace laboratory experiments, biological assays, safety studies or independent experimental validation.
04
Evaluation research
Future Research
Longer-term research directions that extend Spacefield's computational framework into broader molecular and scientific applications.
- Molecular optimisation
- Multi-target discovery
- Protein modelling
- Advanced computational molecular systems
A benchmark result is interpreted within the scope of its dataset, split, endpoint and evaluation protocol.
05
Engineering research
Reproducible scientific software
Research methods are progressively converted into versioned, testable and deployable software so that computational claims can be linked to executable artefacts.
- Stable APIs, batch workflows and command-line tools
- Automated tests and predictor-parity checks
- Dependency control and release packaging
- CSV, JSON, PDF, logs and machine-readable reports
A software release records implementation state; it does not by itself establish scientific or clinical validity.
06
Translational direction
Computational prioritisation
Investigation of how integrated molecular predictions can support candidate comparison and the design of focused experimental programmes.
- Multi-parameter candidate profiles
- Agreement and disagreement across prediction engines
- Target-aware ranking and filtering workflows
- Interfaces for academic and industry collaboration
Computational prioritisation supports decisions about what to test; it does not demonstrate biological efficacy or safety.
Benchmarking and validation
Performance claims require controlled evidence.
Spacefield's evaluation framework is designed to reduce common sources of inflated performance and to preserve a clear boundary between development decisions and final testing.
Development
→
Training partition
Fit candidate models and establish the implementation pipeline without consulting locked test outcomes.
Selection
→
Validation partition
Compare authorised algorithms and choose a frozen approach under a predefined selection plan.
Final assessment
→
Locked test partition
Evaluate the frozen method only after development and selection decisions have been completed.
Generalisation
External evaluation
Where suitable data exist, examine performance on an independent dataset or deployment context.
Credible baselines
New descriptors and models should be compared with appropriate established representations, simple controls and relevant combined systems.
Leakage control
Duplicate molecules, closely related records, target overlap and preprocessing decisions require explicit handling to avoid optimistic estimates.
Multiple metrics
Regression and classification performance should be described through metrics appropriate to the endpoint, data balance and intended use.
Uncertainty and failure
Research reporting should include failed inputs, missing values, out-of-domain cases and uncertainty where the method supports it.
Repeatability
Seeds, environments, dataset fingerprints and execution manifests help determine whether a reported result can be reproduced.
Scope-aware conclusions
Claims should remain bounded by the dataset, model, endpoint, population and evaluation design used in the study.
Research evidence
What accompanies a credible computational result?
A metric is only one part of the evidence. Reviewable research also needs context, provenance and implementation records.
For public studies and software releases, Spacefield aims to connect scientific conclusions with the artefacts required to understand and reproduce the work, subject to dataset licences, confidentiality, intellectual-property protection and responsible disclosure.
Scientific record
Research question, assumptions, endpoint definition, methods and limitations.
Data record
Dataset identity, licence, inclusion criteria, preprocessing and partition fingerprints.
Model record
Descriptor version, algorithm, parameters, training procedure and model hash.
Execution record
Environment, configuration, runtime, warnings, failures and generated artefacts.
Evaluation record
Metrics, baselines, confidence intervals where available and error analysis.
Release record
Version, changelog, integrity manifest, citation information and preservation location.
Publications and research records
Publications, software and scientific records.
Spacefield research outputs are organised so that theoretical work, software, validation evidence and reproducibility records remain distinguishable and citable. Public records will be linked to appropriate repositories and DOI-backed releases as they become available.
Publications
Scientific papers and formal research manuscripts.
- Peer-reviewed journal articles
- Preprints
- Conference papers
Software & Releases
Versioned computational artefacts that make research methods executable and reviewable.
- Versioned software packages
- Platform releases
- SDKs and developer resources
- APIs
Validation & Scientific Records
Evidence and documentation supporting evaluation, traceability and reproducibility.
- Benchmark reports
- Validation studies
- Technical reports
- DOI archives and reproducibility records
Publication status: items described on this page should not be treated as published until a public record, DOI, journal citation or official release link is provided.
Reproducibility and responsible openness
Open where useful, protected where necessary.
Research transparency does not require the indiscriminate publication of confidential code, proprietary models, licensed datasets or security-sensitive deployment details.
Open
Suitable for public release
- Research questions and scientific rationale
- Evaluation methodology and baseline definitions
- Non-confidential benchmark summaries
- Version, citation and integrity information
- Selected examples and public documentation
Controlled
May require protection or restricted access
- Proprietary source code and model artefacts
- Confidential molecular workflows
- Licensed or restricted datasets
- Partner information and unpublished results
- Detailed implementation blueprints relevant to IP
Governed
Shared through appropriate agreements
- Research collaboration materials
- Private validation packages
- Pilot-project data and reports
- Technical due-diligence materials
- Controlled API or software evaluation access
Research collaboration
Connecting computational development with independent expertise.
Spacefield welcomes carefully defined collaborations that can strengthen evaluation, extend scientific scope and connect computational predictions with appropriate experimental evidence.
Continue exploring
From research evidence to versioned software.
Explore the SFT Molecular Platform, review the technology architecture or visit the Version Centre for software and research-release records.