PoC: Improve annotations with DeepFRI

guillecg / mars-biofoundry

Metabolic engineering towards a biofoundry on Mars

MIT License

0 stars 0 forks source link

PoC: Improve annotations with DeepFRI #26

Closed guillecg closed 1 year ago

guillecg commented 1 year ago

Add annotations from DeepFRI and then build the metabolic models again.

[x] Use MSFeature and MSGenome in ModelSEEDpy/code/msgenome.py
[x] Add GO terms from DeepFRI using those classes
[x] Add them to the existing MSGenome
[ ] Use the last part of the ModelSEEDpy pipeline (not the RAST one) for building the metabolic models

guillecg commented 1 year ago

DeepFRI's prediction lack the last digit in the EC number, which overly complicates mapping specific enzymes to a single sequence. For example, a given EC number of third level may map to many enzymes whose substrates are radically different (e.g. EC 1.13.11.-), thus making more difficult to extract specific enzymes from the incomplete annotations.