Scan single mutations · dms

Score every single-residue substitution of a wild-type sequence, as an in-silico deep mutational scan:

adabmDCA dms -d wildtype.fasta -p model/params.dat.gz -o dms

For a sequence of length \(L\) over \(q\) symbols, this produces \(L(q-1)\) mutants (gaps included as a symbol).

Inputs

Option	Meaning
-d, --data	Aligned FASTA with the wild type. If it contains several sequences, the first is used
-p, --path_params	params.dat(.gz) or ptt.h5
-o, --output	Output folder

The wild type must be aligned like the training data: same length, same alphabet.

Outputs

<name>_DMS.fasta, .csv and .json, where <name> is the wild type's FASTA header with non-alphanumeric characters removed. Each mutant is labelled by mutation and score, for example:

>G27A | DCAscore: -0.6

The score is \(\Delta E = E(\text{mutant}) - E(\text{wild type})\). Negative values mean the model prefers the mutant; \(-\Delta E\) is the log of the probability ratio between mutant and wild type. The mutation labels (G27A) use zero-based alignment positions, like every other output of the package; the CSV also gives position_1based. Check the numbering against your assay before joining the two, and remember that alignment positions are not residue numbers of the full-length protein when the alignment has removed insertions.

Interpreting the scores

\(\Delta E\) measures how compatible a substitution is with the family's sequence statistics. It often correlates with measured fitness, but the strength of that correlation depends on the protein, the assay and the selected property; scores should be validated on experimental data before being used as fitness predictions. Substitutions at strongly coupled positions depend on the background: the same mutation can be favourable in one homologue and deleterious in another.

In Python

from adabmDCA import load_model

model = load_model("model/params.dat.gz", alphabet="protein")
scan = model.scan_mutations("MKV...", name="wild_type")
df = scan.to_dataframe()
scan.to_fasta("wild_type_DMS.fasta")

Definition: mutation scores.