When you are exploring a molecular structure, small tasks can interrupt the scientific question you are trying to answer: find a binding site, isolate nearby residues, change a representation, and clear away visual clutter. Each action may be simple on its own, but together they can make a structure harder to navigate—especially when you are still learning the software’s menus and selection syntax.
SAMSON AI offers a way to describe some of these tasks in ordinary language. The practical value is not that every request will be interpreted perfectly; it is that you can turn a structural goal into a sequence of actions, then check each result as you work.
A focused three-step example
Suppose you want to inspect the surroundings of a ligand. You could break the task into three manageable parts:
- Choose the relevant nodes. Ask SAMSON AI to select the ligand, or use a more specific request such as
/select all side chains within 5 angstrom of the ligand. - Apply a useful view. For example, ask it to select the ligand and apply a licorice model:
/do select the ligand and apply a licorice model. - Refine the scene. If surfaces are obscuring the view, a request such as
/do erase all surfaces.can help clear them from the document.
The point is to keep selection and scene changes understandable. A request to select nearby side chains defines what you want to examine; a separate action can then change how relevant molecular nodes appear. If the result is not what you intended, you have a clearer place to revise the request than you would with one long, ambiguous instruction.
Use the right command for the job
The slash commands make it easier to signal what kind of help you want. Use /select when your goal is to identify nodes in the document. SAMSON AI displays the corresponding Node Specification Language (NSL) for the natural-language query and executes it. This is useful when you know the structural idea—such as “all lysines in chain B”—but do not want to construct the selection expression from scratch.
Use /do when you want an action performed in the document. The documented examples include zooming to a binding site, removing water, changing a surface’s color, and applying a model to a ligand. In other words, describe the outcome you want rather than trying to remember the exact menu path.
You can type / in the Assistant input to see the available commands. The command list also includes /script for generating Python scripts with the SAMSON Python API. That is a different route: it can be useful when a task calls for a reusable script or more customized interaction with the document, rather than a one-off natural-language action.
Make requests easier to interpret
Natural language can be ambiguous. A useful habit is to name the object, the operation, and any important constraint. For example, “select all lysines in chain B” is more specific than “select the lysines,” and “select all side chains within 5 angstrom of the ligand” states both the relationship and the distance.
For multi-part work, try one request at a time and inspect the result before moving on. Confirm that the selection corresponds to the intended region and that the visual change helps answer your question. If the result is unclear, make the next request more precise. This is especially important because SAMSON AI can misinterpret prompts or provide inaccurate information; critical scientific conclusions should be checked rather than inferred from an AI response alone.
This workflow is most helpful for reducing friction while you explore a structure—not for replacing your judgment about what the structure means. You remain responsible for deciding whether a selection, representation, or interpretation is appropriate for your analysis.
To see the full command list, more examples, and the documented limitations, read the SAMSON AI documentation.
Note: SAMSON and all SAMSON Extensions are free for non-commercial use. Get SAMSON.