Filling in a tool and collecting its results
How to set file and text inputs, check a file before using it, find a file you don't have yet, and collect results from runs that take time.
Filling in a tool and collecting its results
Every tool under Model works the same way: fill in the form, run it, and collect what it produced. This page covers the parts of that loop that are easy to get stuck on.
Two kinds of input
A tool form has two kinds of row.
File inputs are always a dropdown of files already saved in this project. You cannot upload from your desktop in the middle of a run, and that is deliberate: a run has to point at a real evidence row with a source and a history, or the provenance it records would be a fiction. The dropdown only offers files of the types that particular input accepts, so a tool that needs coordinates will not offer you a spreadsheet.
Text inputs are everything else — a number, a short piece of text, a choice from a list, or a yes/no switch. Most have a sensible default already filled in, and a hint underneath saying what the value does. If you are not sure, leave the default: the defaults are chosen to be the ordinary case.
Checking a file before you use it
Next to a selected file you will find File info ▸. Open it and RxCubed reads the file itself and tells you what is actually inside — for a structure, the chains, ligands, and models it contains; for other types, whatever that format makes available.
This is worth a click before every run, because it answers the question that causes most confusing results: is this the file I think it is? A structure with no ligand will not produce contacts. A model with one chain cannot be aligned to two. File info ▸ shows you that in a second, without opening the 3D viewer.
▸ Don't have the file?
If the dropdown is empty, or does not contain what you need, the picker carries a ▸ Don't have the file? toggle underneath it.
Opening it lists the searches that produce this exact file type — so a coordinates input points you at RCSB PDB and AlphaFold DB, a sequence input points you at UniProt, and so on. Each opens in a new tab so you keep your half-filled form. The list ends with the manual upload page, for a file that came from somewhere RxCubed does not connect to.
The list is built from what each connector says it can download, so it cannot drift out of step with what those searches actually return.
What a run produces
A successful run writes one evidence row per output file. Some tools produce several — the contact mapper writes both a TSV and a JSON, for example — and each becomes its own row under the same run, so each can be opened, downloaded, and cited on its own.
Alongside the files, a run records a summary and often a table, shown in the Result panel directly under the form as soon as the run finishes. That is the part to read first; the files are for when you need the detail or want to feed another tool.
A run that fails writes nothing at all. There is no half-saved output and no empty row to tidy up — a failed run must never end up in the scientific record.
Runs that take time: the Run Results tab
Some tools finish immediately and show their result under the form. Others — molecular dynamics, docking, folding, and anything sent to Neurosnap — take minutes or hours. Those are asynchronous runs, and they have their own tab.
Tool Selector has three tabs: Run, Run Results, and All Runs. Run Results is where a job that is still working, or has finished but not yet been brought back, waits for you.
- Pending Runs — the job is still being worked on. Nothing to do but wait.
- Results not collected — the job has finished and its results are ready, but they are not in your project yet.
This second list matters: a finished job's files are not saved until you collect them. Until you do, the run will not appear properly in Results Hub and its outputs cannot be used as inputs to anything else.
Nothing on this tab refreshes by itself. For a Neurosnap job that is on purpose — every status check spends your own Neurosnap allowance, so RxCubed never checks on your behalf. Open the tab and press the check button when you are ready, then collect anything that has finished.
A tool that runs locally and finishes immediately never appears in either list. There is nothing left to collect.
What to do next
Once outputs are collected they behave like any other evidence: open them from Results Hub, inspect them in the 3D viewer where that applies, feed them into another tool, or pull them into a report. The Graph tab in Results Hub shows which inputs produced them and what they went on to feed.