Molecular Biological Similarity with Conditional Variational Autoencoder

Modeling the molecular biological similarity with conditional variational autoencoder. The model is trained on the ChEMBL dataset using BioBricks.

A presentation of this work is available on google drive ChemHarmony and Biosim

Development

1. install docker and nvidia-docker2
2. test nvidia-docker:
   docker run --rm --gpus all nvidia/cuda:11.7.1-devel-ubuntu20.04 nvidia-smi
3. build dockerfile
4. run dockerfile:
   docker run -p 6515:6515 -v .:/chemsim --rm --gpus all -it --name chemsim biobricks-ai/cvae

Evaluation

We want to evaluate this model using the same benchmarks from the molformer paper and using some new benchmarks for the NIEHS acute inhalation project and Tox24:

1. SIDER
2. Tox21
3. ClinTox
4. HIV Activity
5. BACE
6. CoMPAIT - Collaborative Modeling Project for Acute Inhalational Toxicity

and for regression tasks

1. QM9 (quantum mechanical properties)
2. ESOL (solubility from moleculenet)
3. FreeeSolv (free solvation energy from moleculenet?)
4. lipophilicity (lipophilicity from moleculenet)
5. Tox24

Model	BBBP	Tox21	ClinTox	HIV	BACE	SIDER
RF	71.4	76.9	71.3	78.1	86.7	68.4
SVM	72.9	81.8	66.9	79.2	86.2	68.2
MGCN	85.0	70.7	63.4	73.8	73.4	55.2
D-MPNN	71.2	68.9	90.5	75.0	85.3	63.2
DimeNet	-	78.0	76.0	-	-	61.5
Hu et al.	70.8	78.7	78.9	80.2	85.9	65.2
N-gram	91.2	76.9	85.5	83.0	87.6	63.2
MolCLR	73.6	79.8	93.2	80.6	89.0	68.0
GraphMVP-C	72.4	74.4	77.5	77.0	81.2	63.9
GeomGCL	-	85.0	91.9	-	-	64.8
GEM	72.4	78.1	90.1	80.6	85.6	67.2
ChemBERTa	64.3	-	90.6	62.2	-	-
MoLFormer-XL	93.7	84.7	94.8	82.2	88.21	69.0

• Bold indicates the top-performing model.
• '—' signifies that the values were not reported for the corresponding task.

Model	QM9 (MAE)	QM8 (MAE)	ESOL (RMSE)	FreeSolv (RMSE)	Lipophilicity (RMSE)
GC	4.3536	0.0148	0.970	1.40	0.655
A-FP	2.6355	0.0282	0.5030	0.736	0.578
MPNN	3.1898	0.0143	0.58	1.150	0.7190
MoLFormer-XL	1.5894	0.0102	0.2787	0.2308	0.5289

• Bold indicates the top-performing model.

Run service

docker run -p 6515:6515 -v .:/chemsim --rm --gpus all -it --name chemsim biobricks-ai/cvae

curl -X GET "http://localhost:6515/predict?property_token=5042&inchi=InChI=1S/C9H8O4/c1-6(10)13-8-5-3-2-4-7(8)9(11)12/h2-5H,1H3,(H,11,12)"

ssh -Nf -R 12000:localhost:6515 ubuntu@api.insilica.co

curl -X GET "https://api.insilica.co/service/run/chemsim/predict?property_token=5042&inchi=InChI=1S/C9H8O4/c1-6(10)13-8-5-3-2-4-7(8)9(11)12/h2-5H,1H3,(H,11,12)"

References

• Conditional Variational Autoencoder for Molecular Biological Similarity

License

MIT