From “Where’s That Selection?” to a Clear Molecular Workflow

In molecular modeling, the science is often not the slow part. It is the sequence of small interface tasks around it: finding the right atoms, changing a representation, clearing a surface, or repeating an operation after adjusting a selection. A useful shortcut is to describe the task in plain language and let SAMSON AI translate it into an action or selection.

SAMSON AI works inside SAMSON and can answer questions about commands, menus, editors, apps, selections, and scripts. You can open it from Interface > Assistant, or use Ctrl+0 on Windows and Linux, and Cmd+0 on Mac. Its commands are most useful when you match the command to the kind of task you want to perform.

Use the right command for the job

For an action in the current document, use /do. For example, you can ask it to zoom to a binding site, remove water, erase surfaces, or select a ligand and apply a licorice model. This is useful when you know the outcome you want but do not want to search through menus to locate the corresponding operation.

For a selection described by its molecular properties or spatial relationship, use /select. It can interpret requests such as /select all side chains within 5 angstrom of the ligand. or /select all lysines in chain B. SAMSON AI displays the corresponding Node Specification Language (NSL) expression and executes it. Seeing the expression gives you a chance to understand how the natural-language request was interpreted, rather than treating the selection as an unexplained result.

A practical selection-to-action sequence

1. Describe the target precisely. Include a chain, molecule, distance, or other distinguishing detail when it matters. For instance: /select all side chains within 5 angstrom of the ligand.

2. Inspect the resulting selection. Check that the selected nodes match the intended region before building further work on them. Natural-language requests can be ambiguous, and the assistant may misinterpret unclear input.

3. Apply the next operation. Once the target is right, request an action such as changing a model or zooming to the selected region. You can also make the request directly with /do, for example: /do select the ligand and apply a licorice model.

4. Check the document after the action. Confirm that the visible result is what you expected. This is especially important before relying on a selection or change in a later modeling step.

The benefit of this approach is not that every task becomes automatic. It is that you can express many routine operations in the language of the modeling problem, then review what the assistant selected or changed. Specific requests are easier to evaluate than vague ones: “select the binding site” may be useful, while adding a relevant molecule, chain, or distance can make your intent clearer.

When the task needs more than an action

If the workflow involves custom calculations or a repeatable procedure, try /script. It generates Python scripts that use the SAMSON Python API and can be combined with other Python modules. Examples include translating atoms, creating a small GUI for a movement distance, or computing the gyration radius of a receptor. Review generated scripts before using them as part of a critical workflow.

If your question depends on a document, /learn can read a regular web page or PDF from a URL. You can then ask about the learned material with /refer. Learned information persists until you use /forget, which clears it. This can be handy when you want to consult a paper or reference without leaving the modeling workflow, but check important answers against trusted sources: the assistant may be incomplete or inaccurate.

SAMSON AI is based on GPT-5 in the Professional plan. As with other AI tools, use clear prompts and independently verify critical scientific information and document changes.

For command details and examples, see the SAMSON AI documentation.

Note: SAMSON and all SAMSON Extensions are free for non-commercial use. Get SAMSON.