Insilico Medicine and DeltaWaveOS are often compared as AI platforms for molecular discovery. That comparison changes in cosmetics, agritech, and specialty chemistry, where the objective is not a therapeutic target and clinical programme. Formulation compatibility, sensitisation, biodegradability, environmental fate, crop selectivity, and regional compliance can become first-class optimisation criteria. The architecture must be able to carry that context through the same design, evaluation, and evidence loop.

Insilico Medicine: deep pharmaceutical grounding

Insilico’s public product architecture connects PandaOmics for target and multi-omics analysis with Chemistry42 for generative small-molecule design and optimisation [1]. Its strongest public proof is pharmaceutical: target discovery, medicinal chemistry, preclinical work, and clinical translation. That is a genuine advantage for therapeutic programmes.

It would be inaccurate to say Insilico has no activity outside pharma. Its 2025 listing document describes collaborations in advanced materials, agriculture, nutrition, and other discovery work [2]. The distinction is architectural emphasis: the public software suite, scientific vocabulary, and most visible validation remain centred on drug discovery.

DeltaWaveOS: scientific objectives as configurable context

DeltaWaveOS treats industry-specific constraints as programme context rather than a new product silo. EVE can coordinate generative design, property prediction, literature review, scientific tools, and multi-objective ranking; the customer can add approved models, internal data, and custom reward functions for the programme. An agritech workflow can reward potency and environmental profile, while a cosmetics workflow can prioritise stability, exposure, and formulation constraints.

This does not make the underlying scientific domains interchangeable. Domain experts, validated assays, and appropriate models remain necessary. The benefit is that the orchestration, provenance, deployment, and learning infrastructure does not have to be rebuilt when the objective changes.

Compliance is part of the workflow, not a final checkbox

Non-pharmaceutical programmes have their own regulatory and industry frameworks. REACH governs chemical registration and risk management in the European Union, while IFRA Standards provide a risk-management system for fragrance ingredients [3], [4]. A useful AI system therefore needs to preserve which evidence, model, and rule informed a decision. Compliance cannot be represented by a generic “safe” score detached from the source and programme context.

Deployment boundaries matter for trade-secret science

Formulations, assay results, and candidate structures may remain confidential rather than enter a patent disclosure. The European Commission’s data strategy now explicitly links trusted data flows with data sovereignty [5]. DeltaWaveOS can be deployed in private cloud, on-premises, or air-gapped environments, with model and tool access governed inside the chosen customer boundary. This is a deployment capability, not a claim that every configuration automatically stays in one jurisdiction.

PCP supplies the portable compute layer underneath EVE and is being developed for heterogeneous, distributed hardware through SPRIND’s Composite Learning Challenge [7]. The same boundary can contain open models, customer-specific reinforcement-learning environments, workflow artifacts, and experimental results.

Which platform fits which programme?

Insilico Medicine is a strong candidate when the centre of gravity is pharmaceutical target discovery and clinical translation. DeltaWaveOS becomes more relevant when one organisation wants to operate several molecular programmes through a common agentic and auditable system, retain model choice, and control where data and execution live. Its auditable agent research shows how tool calls and molecular lineage can remain inspectable across an optimisation run [6].

The right evaluation is a programme-level pilot: use representative proprietary data, explicit success criteria, the required regulatory sources, and the intended deployment boundary. For a broader market view, read the 2026 molecular discovery software guide.

References

  1. Insilico Medicine. Pharma.AI platform and product index. insilico.com/llmstxt
  2. Insilico Medicine Cayman TopCo. (2025). Global offering document, including other discovery collaborations. ir.insilico.com/global-offering-en.pdf
  3. European Chemicals Agency. Understanding REACH. echa.europa.eu/understanding-reach
  4. International Fragrance Association. Introduction to the IFRA Standards. ifrafragrance.org/ifra-standards
  5. European Commission. (2026). European Data Union Strategy. digital-strategy.ec.europa.eu/data-union
  6. Ünlü, A., Rohr, P., & Celebi, A. (2025). An Auditable Agent Platform For Automated Molecular Optimisation. arXiv:2508.03444
  7. SPRIND. Composite Learning Challenge. sprind.org/composite-learning