liveBook · Manning

Fusion

As explained in Build AI Drug Discovery Pipelines

Data fusion is the integration of information from multiple data modalities into a shared representation or combined inference process.

Data fusion combines distinct data types in one learned representation. In drug discovery, this differs from orchestration, where separate single-modality models pass outputs through a pipeline, and from end-to-end integration of discovery stages. section 13.1.2 section 13.0

Pipeline orchestration is not fusion

A protein-sequence embedding can inform structure prediction and function annotation, guide molecular generation, and receive feedback from cellular-assay images. Because separate models pass outputs along, this is orchestration rather than genuine fusion. LucaOne and ProCyon are identified as closer examples of genuine fusion. section 13.1.2

How Build AI Drug Discovery Pipelines covers it

How deep it goes
Defines genuine fusion as distinct data types sharing one learned representation and contrasts it with model orchestration.
Examples
Uses protein-sequence embeddings passed through prediction, generation, and assay-feedback steps to illustrate orchestration rather than fusion.
What you can do after
Evaluate whether a drug-discovery system fuses modalities in a shared representation or passes outputs between separate models.
How it connects
Connects multimodal AI with protein embeddings, structure prediction, function annotation, and molecular generation.

Also explained in

Compare how 2 books explain fusion →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Build AI Drug Discovery Pipelines. The section links point to the parts of the book it draws on.