Finding the right atoms or residues can interrupt an otherwise straightforward modeling task. You may know what you want—such as lysines in one chain or side chains near a ligand—but translating that idea into a precise selection takes time. In SAMSON, you can describe the selection in natural language and have SAMSON AI show and execute the corresponding Node Specification Language (NSL) query.
/ in the Assistant input to see available commands.Make the selection specific
Open the Assistant from Interface > Assistant, or use Ctrl+0 on Windows and Linux, or Cmd+0 on Mac. Then start a request with /select. For example:
/select all lysines in chain B./select all side chains within 5 angstrom of the ligand./select the binding site.
The useful detail is that the assistant displays the NSL query it generated and executes it. That gives you more than a selected region: you can also inspect the selection logic. If the result is not what you intended, refine the wording—make the chain, distance, or molecular feature explicit—and check the updated query and selection before using it in a later step.
Separate “what to select” from “what to do”
A practical way to avoid muddled requests is to break a task into two decisions. First, specify the target with /select. Then describe the action with /do. The documentation gives examples such as /do select the ligand and apply a licorice model. and /do select the gaussian surface and make it light blue. You can also ask SAMSON AI to zoom to the binding site or remove water.
This separation makes it easier to catch a mistake early: confirm that the intended nodes are selected before asking for a visual or structural change. Use concrete language, especially when terms like “site,” “near,” or “surface” could refer to more than one thing in a complex document. The assistant can misinterpret an ambiguous request, so treat the displayed result as something to review, not as a substitute for your judgment.
When a script is a better fit
For a one-off selection or document action, a natural-language command may be enough. If you need a custom procedure or want to work with the document through Python, use /script to request a script based on the SAMSON Python API. Documented examples include translating atoms, creating a small GUI for a movement operation, and computing a receptor’s gyration radius. Generated code should still be reviewed and tested for your specific document and task.
A useful habit for AI-assisted modeling
Keep requests narrow, inspect the selection or generated query, and verify important results in the molecular document. SAMSON AI can be helpful for navigating commands and turning plain-language intent into actions, but it can produce inaccurate or misinterpreted results. For critical work, cross-check the outcome rather than relying on the wording of a response alone.
For command details and other SAMSON AI capabilities, including learning from URLs and asking questions about learned material, see the SAMSON AI documentation.
Note: SAMSON and all SAMSON Extensions are free for non-commercial use. Get SAMSON.