Data for mutation effects on protein stability.
In molecular biology, we call an amino acid in a protein sequence a residue.

These data were collected initially for my degree thesis research, which aimed to build a model for ∆∆G prediction from Single Sequence because at that time we did not have the AlphaFold2. Now I already gave up that project and changed my thesis from pure dry-lab to some "real" protein design project, and developed DDGScan by the way.

For train.csv and test.csv:
PDB: PDB ID code, get widetype 3D structure from https://files.rcsb.org/download/{pdb_id}.pdb.
wildtype: Wildtype amino acid.
position: Residue number of mutation in sequence.
mutation: Changed amino acid in mutant_seq.
ddG: Experimentally measured ∆∆G(folding), positive means more stable.
sequence: Wildtype protein sequence.
mutant_seq: Mutated protein sequence.

For tm.csv:
PDB: PDB ID code, get widetype 3D structure from https://files.rcsb.org/download/{pdb_id}.pdb.
WT: Wildtype amino acid.
position: Residue number of mutation in sequence.
MUT: Changed amino acid in mutant_seq.
dTm: Experimentally measured ∆T(melting), positive means more stable.
sequence: Wildtype protein sequence.
mutant_seq: Mutated protein sequence.

The V1 data was first released for kaggle competition novozymes-enzyme-stability-prediction, I cleaned and updated these data to this version, hope this will be helpful.

pdb: PDB ID code, get widetype 3D structure from https://files.rcsb.org/download/{pdb_id}.pdb.
wildtype: Wildtype amino acid.
pdb_resseq: Auth. Resseq number in the 6th col in a pdb file. Not always starts from 1!
seq_index: Index in the seq string where a single mutation happens starts from 0.
mutation: Changed amino acid in mut_seq.
wt_seq: Wildtype protein sequence.
mut_seq: Wildtype protein sequence.
ddG: Experimentally measured ∆∆G(folding), positive means more stable.
group: For K-fold CV. The test.csv does not have this.

Sorry for the inconsistent naming of columns. Time has changed me.
The data stored outside v1 and v2 is kept for some kaggle notebooks that already used this data.

A tool designed for enzyme stability prediction: DDGscan