Entities
Entity note는 molecular AI와 protein modeling 전반에 등장하는 object를 정의합니다.
Entity 자체는 research area가 아닙니다. Task, representation, model, label, split, metric이 작동하는 대상입니다.
The basic pattern is:
where is an entity, is context such as target, assay, pocket, source, or time, is a representation function, and is the model output.
Molecular Modeling에서는 같은 entity도 서로 다른 역할을 할 수 있습니다.
| Entity | Common role | Typical risk |
|---|---|---|
| Molecule | input, ligand, generated object | scaffold leakage or inconsistent standardization |
| Protein | target, sequence, structure, context | homolog leakage or missing biological context |
| local structure context | ligand-defined pocket leakage | |
| Complex | pose, interaction, affinity example | mixing pose quality with binding affinity |
| Assay | measurement context | incompatible labels treated as comparable |
| Dataset | collection and split boundary | row-level split hiding entity overlap |
Map
Biological and Chemical Objects
Representations and Data
Checks
- What is one example: molecule, protein, pocket, complex, assay row, sequence region, or dataset record?
- What is the model-ready representation of the entity?
- What context changes the entity’s meaning?
- What label, if any, is attached to the entity?
- What split unit prevents the same or nearly same entity from leaking across train and test?
- Does the note describe the object itself, the way to represent it, or the task performed on it?