If you work with molecular structures, small tasks can interrupt a larger train of thought: find a binding site, select nearby residues, change a representation, or remove waters. The challenge is not always knowing what you want to do; it is remembering the right menu, selection syntax, or sequence of commands to do it.
SAMSON AI offers an in-application way to describe some of these tasks in natural language. You can open it from Interface > Assistant, or use Ctrl+0 on Windows and Linux and Cmd+0 on Mac. Its commands are useful for different kinds of work: /select for choosing nodes, /do for actions in the document, and /script for generating Python scripts.
/ in the Assistant input to see the available commands.Start with the task, not the menu
For a structural selection, describe both the target and any useful constraint. For example, /select all side chains within 5 angstrom of the ligand. SAMSON AI translates the natural-language request into Node Specification Language (NSL), displays the corresponding expression, and executes it. That makes the command useful both as a shortcut and as a way to encounter NSL in the context of a real selection.
Other documented examples include selecting the binding site or selecting lysines in chain B. The more precisely you name the object, chain, or distance you care about, the easier it is to judge whether the result matches your intention. After the selection runs, inspect what is highlighted before building on it—especially when a selection will determine the next modeling step.
Separate selecting from acting
A practical way to make a request easier to check is to break it into two parts: first identify the relevant nodes, then ask for an action. For instance, select the ligand and apply a licorice model with /do select the ligand and apply a licorice model. Other examples from the documentation include zooming to the binding site, erasing surfaces, removing water, and changing a Gaussian surface to light blue.
This distinction helps when a request combines “what should I work on?” with “what should happen to it?” Use /select when the central task is identifying nodes and you want to see the resulting NSL. Use /do when you want SAMSON AI to carry out an action in the document. In either case, check the document afterward; AI-generated results can be inaccurate or misinterpret an ambiguous request.
Use scripts when the task needs custom logic
Some jobs are better expressed as a small tool or calculation than as a single document action. The /script command generates Python scripts that use the SAMSON Python API to interact with the document. Examples include translating all atoms by one angstrom along the z direction, creating a small distance-entry interface with a Move button, or computing a receptor’s radius of gyration.
Before asking for a script, state the object to operate on, the operation, and important units or direction. Then review the generated code and confirm that it matches the intended calculation or transformation before relying on its output. SAMSON AI can also be asked to learn from a regular webpage or PDF using /learn, answer questions about that material using /refer, and clear learned information using /forget. Learned information persists across SAMSON sessions until it is forgotten, so use these commands deliberately.
SAMSON AI is based on GPT-5 in the Professional plan. Like other AI tools, it can misunderstand requests, omit recent information, or produce inaccurate answers. Treat its response as assistance rather than a substitute for checking selections, scripts, and scientific conclusions.
For command details and examples, see the SAMSON AI documentation.
Note: SAMSON and all SAMSON Extensions are free for non-commercial use. Get SAMSON.