Basic Workflow Tutorial¶
This tutorial shows the simplest end-to-end Surfacia workflow from input structures to interpretable output.
Goal¶
Learn how to:
prepare a minimal input table
run the main workflow command
locate the most important output files
understand what each stage produced
Before You Start¶
You should already have:
Surfacia installed
Gaussian and Multiwfn available in your environment
a CSV file with at least
Sample NameandSMILEScolumns
Minimal Input Example¶
Sample Name,SMILES
caffeine,CN1C=NC2=C1C(=O)N(C(=O)N2C)C
aspirin,CC(=O)OC1=CC=CC=C1C(=O)O
ibuprofen,CC(C)CC1=CC=C(C=C1)C(C)C(=O)O
Recommended Path¶
Prepare your molecule table and target values.
Run the standard workflow command.
Inspect generated descriptor tables.
Review the compact model and SHAP outputs.
Run the Workflow¶
For a first pass, use the full workflow with resume enabled:
surfacia workflow -i molecules.csv --resume --test-samples "1,2"
What to Inspect First¶
After the workflow finishes, focus on these outputs first:
the complete descriptor table
the training or modeling table used downstream
the compact retained feature list
the SHAP visualization outputs
Practical Reading Order¶
If you are new to Surfacia, this order usually works best:
confirm the workflow finished cleanly
inspect whether descriptor columns look reasonable
check which features survived compact modeling
inspect SHAP outputs only after the retained features make chemical sense
What a Good First Result Looks Like¶
A good first tutorial run usually gives you:
a clean output directory structure
descriptors spanning size, shape, electronics, and surface analysis
a compact feature subset that is smaller than the raw matrix
SHAP results that can be connected back to recognizable chemistry
Common Beginner Mistakes¶
starting with too many molecules before checking one small test run
trusting prediction metrics without checking the retained descriptors
treating all features as equally meaningful instead of focusing on the compact model
using SHAP before confirming the feature table itself is sensible
What to Do Next¶
After one successful run, choose a direction:
go to Advanced Analysis Tutorial if you want to compare Surfacia modes
go to Machine Learning Tutorial if you want to focus on compact modeling
go to Examples if you want problem-oriented usage patterns