/deepprot

machine learning for protein engineering

Apache License 2.0Apache-2.0

DeepProt

A simple, universal API for machine learning on protein engineering.

Quick Start

Train a protein model in four lines of Python.

import deepprot

antibodies = dp.data.Dataset("cancer_cure.csv", X="seq", y="targetA")
dl = dp.data.DataLoader(antibodies, featurizer=deepprot.feat.KideraFactors())

cnn = dp.model.CNN()
cnn.fit(dl)

Use DeepProt's rich library of off-the-shelf models and featurizers:

# Dozens of featurization schemes...
dl = dp.data.DataLoader(solubility_data, featurizer=deepprot.feat.ProtBert())
dl = dp.data.DataLoader(new_exp, featurizer=deepprot.feat.BLOSUMIndices())

# Generic architectures that resize to variable input under the hood.
model = dp.models.MLP()
model = dp.models.Transformer()

Getting Involved

  • Slack: For discussions about development, questions about usage, and feature requests.
  • GitHub Issues: For reporting bugs.
  • kenny@latch.ai: For questions about how to use DeepProt.