update airtable csv [skip ci]

.github/scripts/update_metadata.py

@@ -103,7 +103,13 @@ def populate_metadata(self):
  if self.metadata["Title"] == "":
  self.metadata["Title"] = self.json_input["model_name"]
  if self.metadata["Description"] == "":
- self.metadata["Description"] = self.json_input["model_description"]
+ # Check if model_description is a list
+ if isinstance(self.json_input["model_description"], list):
+ # Join the list elements into a single string separated by commas
+ self.metadata["Description"] = ", ".join(self.json_input["model_description"])
+ else:
+ # If it's already a string, just assign it directly
+ self.metadata["Description"] = self.json_input["model_description"] 
  if self.metadata["Publication"] == "":
  self.metadata["Publication"] = self.json_input["publication"]
  if self.metadata["Source Code"] == "":

ersilia/hub/content/data/models.tsv

@@ -37,7 +37,7 @@ eos9ueu ['Compound'] Online https://github.com/ersilia-os/eos9ueu https://www.nc
 $ ersilia api -i 'CCCOCCC'
 $ ersilia close" 2023-11-01 https://github.com/miquelduranfrigola
 eos481p ['Compound'] Pretrained https://github.com/ersilia-os/eos481p https://papers.nips.cc/paper/2020/hash/94aef38441efa3380a3bed3faf1f9d5d-Abstract.html https://github.com/tencent-ailab/grover MIT ['Probability'] "Prediction across the ToxCast toxicity panel, containing hundreds of toxicity outcomes, as part of the MoleculeNet benchmark. This model has been trained using the GROVER transformer (see eos7w6n or grover-embedding for a detail of the molecular featurization step with GROVER)
-" Ready grover-toxcast ToxCast toxicity panel ['Toxicity', 'ToxCast', 'Chemical graph model'] Single ['eos481p', 'grover-toxcast'] Probability of toxicity against 617 biological targets ['Classification'] Amna-28 List ['Float'] https://hub.docker.com/r/ersiliaos/eos481p ['AMD64', 'ARM64'] https://ersilia-models-zipped.s3.eu-central-1.amazonaws.com/eos481p.zip {'label': 'GitHub', 'url': 'https://github.com/ersilia-os/eos481p'} "$ ersilia serve grover-toxcast
+" Ready grover-toxcast ToxCast toxicity panel ['Toxicity', 'ToxCast', 'Chemical graph model'] Single ['eos481p', 'grover-toxcast'] Probability of toxicity against 617 biological targets ['Classification'] Amna-28 List ['Float'] https://hub.docker.com/r/ersiliaos/eos481p ['AMD64'] https://ersilia-models-zipped.s3.eu-central-1.amazonaws.com/eos481p.zip {'label': 'GitHub', 'url': 'https://github.com/ersilia-os/eos481p'} "$ ersilia serve grover-toxcast
 $ ersilia api -i 'CCCOCCC'
 $ ersilia close" 2022-07-13 https://github.com/Amna-28
 eos4tcc ['Compound'] Pretrained https://github.com/ersilia-os/eos4tcc https://academic.oup.com/bib/article-abstract/23/4/bbac211/6609519 https://github.com/GIST-CSBL/BayeshERG GPL-3.0 ['Probability'] "BayeshERG is a predictor of small molecule-induced blockade of the hERG ion channel. To increase its predictive power, the authors pretrained a bayesian graph neural network with 300,000 molecules as a transfer learning exercise. The pretraining set was obtained from Du et al, 2015, and the fine tuning dataset is a collection of 14,322 molecules from public databases (8488 positives and 5834 negatives). The model was validated on external datasets and experimentally, from 12 selected compounds (>0.95 probability) one candidate showed strong hERG inhibition (IC 50 < 1 μM) and three moderate (1 μM < IC 50 < 10 μM) in a patch-clamp in vitro assay.
@@ -59,7 +59,8 @@ eos4e40 ['Compound'] Pretrained https://github.com/ersilia-os/eos4e40 https://pu
 " Ready chemprop-antibiotic Broad spectrum antibiotic activity ['E.coli', 'IC50', 'Antimicrobial activity', 'Chemical graph model'] Single ['eos4e40', 'chemprop-antibiotic'] Probability that a compound inhibits E.coli growth. The inhibition threshold was set at 80% growth inhibition in the training set. ['Classification'] miquelduranfrigola Single ['Float'] https://hub.docker.com/r/ersiliaos/eos4e40 ['AMD64', 'ARM64'] https://ersilia-models-zipped.s3.eu-central-1.amazonaws.com/eos4e40.zip {'label': 'GitHub', 'url': 'https://github.com/ersilia-os/eos4e40'} "$ ersilia serve chemprop-antibiotic
 $ ersilia api -i 'CCCOCCC'
 $ ersilia close" 2018-06-06 https://github.com/miquelduranfrigola
-eos8bhe https://github.com/ersilia-os/eos8bhe https://arxiv.org/pdf/2310.10773.pdf https://github.com/datamol-io/safe/tree/main In progress datamol-io-safe safe Inyrkz {'label': 'GitHub', 'url': 'https://github.com/ersilia-os/eos8bhe'} "$ ersilia serve datamol-io-safe
+eos8bhe ['Compound'] Pretrained https://github.com/ersilia-os/eos8bhe https://arxiv.org/pdf/2310.10773.pdf https://github.com/datamol-io/safe/tree/main CC ['Compound'] "The context discusses a novel notation system called Sequential Attachment-based Fragment Embedding (SAFE) that improves upon traditional molecular string representations like SMILES. SAFE reframes SMILES strings as an unordered sequence of interconnected fragment blocks while maintaining compatibility with existing SMILES parsers. This streamlines complex molecular design tasks by facilitating autoregressive generation under various constraints. The effectiveness of SAFE is demonstrated by training a GPT2-like model on a dataset of 1.1 billion SAFE representations that exhibited versatile and robust optimization performance for molecular design.
+" In progress scaffold-morphing Scaffold Morphing ['Compound generation'] Single Model generates new molecules from input molecule by replacing core structures of input molecule. ['Generative'] Inyrkz List ['String'] https://hub.docker.com/r/ersiliaos/eos8bhe ['AMD64'] https://ersilia-models-zipped.s3.eu-central-1.amazonaws.com/eos8bhe.zip {'label': 'GitHub', 'url': 'https://github.com/ersilia-os/eos8bhe'} "$ ersilia serve scaffold-morphing
 $ ersilia api -i 'CCCOCCC'
 $ ersilia close" 2024-01-12 https://github.com/Inyrkz
 eos44zp ['Compound'] Pretrained https://github.com/ersilia-os/eos44zp https://dmd.aspetjournals.org/content/49/9/822 https://github.com/ncats/ncats-adme None ['Probability'] "Analysis of metabolic stability, determining the inhibition of CYP450 activity and whether the compounds are a substrate for the CYP450 enzymes. The data to build these models is publicly available at PubChem, AID1645840, AID1645841, AID1645842. The tested cyps include CYP2C9, CYP2D6 and CYP3A4.
@@ -136,6 +137,9 @@ eos8a5g ['Compound'] Pretrained https://github.com/ersilia-os/eos8a5g https://gi
 " Ready molbloom MolBloom: molecule purchasability in ZINC20 ['ZINC', 'Compound generation'] Single ['eos8a5g', 'molbloom'] It returns a boolean (True/False) suggesting whether the molecule is commercially available or not. ['Classification'] Amna-28 Single ['String'] https://hub.docker.com/r/ersiliaos/eos8a5g ['AMD64', 'ARM64'] https://ersilia-models-zipped.s3.eu-central-1.amazonaws.com/eos8a5g.zip {'label': 'GitHub', 'url': 'https://github.com/ersilia-os/eos8a5g'} "$ ersilia serve molbloom
 $ ersilia api -i 'CCCOCCC'
 $ ersilia close" 2022-11-02 https://github.com/Amna-28
+eos2401 https://github.com/ersilia-os/eos2401 https://arxiv.org/pdf/2310.10773 https://github.com/datamol-io/safe In progress scaffold-decoration Scaffold Decoration Inyrkz {'label': 'GitHub', 'url': 'https://github.com/ersilia-os/eos2401'} "$ ersilia serve scaffold-decoration
+$ ersilia api -i 'CCCOCCC'
+$ ersilia close" 2024-02-20 https://github.com/Inyrkz
 eos7qga ['Compound'] Pretrained https://github.com/ersilia-os/eos7qga https://doc.datamol.io/stable/tutorials/Preprocessing.html https://github.com/datamol-org/datamol Apache-2.0 ['Compound'] "Using the Datamol package, the model receives a SMILE as input, then goes through a process of sanitizing and standardization of the molecule to generate four outputs: Canonical SMILES, SELFIES, InChI and InChIKey
 " Ready datamol-smiles2canonical Converter of SMILES in Canonical, Selfie, Inchi, Inchi Key form ['Chemical notation'] Single Compound represented in its canonical SMILES, SELFIES, InChI and InChIKey forms ['Representation'] carcablop Matrix ['String'] https://hub.docker.com/r/ersiliaos/eos7qga ['AMD64', 'ARM64'] https://ersilia-models-zipped.s3.eu-central-1.amazonaws.com/eos7qga.zip {'label': 'GitHub', 'url': 'https://github.com/ersilia-os/eos7qga'} "$ ersilia serve datamol-smiles2canonical
 $ ersilia api -i 'CCCOCCC'
@@ -347,7 +351,7 @@ eos7jlv ['Compound'] Online https://github.com/ersilia-os/eos7jlv https://online
 $ ersilia api -i 'CCCOCCC'
 $ ersilia close" 2022-08-18 https://github.com/Amna-28
 eos30f3 ['Compound'] Pretrained https://github.com/ersilia-os/eos30f3 https://pubs.rsc.org/en/content/articlehtml/2022/ra/d1ra07956e https://github.com/AI-amateur/DMPNN-hERG None ['Score'] "This model leverages the ChemProp network (D-MPNN, see original Stokes et al, Cell, 2020 for more information) to build a predictor of hERG-mediated cardiotoxicity. The model has been trained using a dataset published by Cai et al, J Chem Inf Model, 2019, which contains 7889 molecules with several cut-offs for hERG blocking activity. The authors select a 10 uM cut-off. This implementation of the model does not use any specific featurizer, though the authors suggest the moe206 descriptors (closed-source) improve performance even further.
-" Ready dmpnn-herg Prediction of hERG Channel Blockers with Directed Message Passing Neural Networks ['Cardiotoxicity', 'hERG', 'Toxicity'] Single Probability of blocking hERG (cut-off: 10uM) ['Classification'] leilayesufu Single ['Float'] https://hub.docker.com/r/ersiliaos/eos30f3 ['AMD64'] https://ersilia-models-zipped.s3.eu-central-1.amazonaws.com/eos30f3.zip {'label': 'GitHub', 'url': 'https://github.com/ersilia-os/eos30f3'} "$ ersilia serve dmpnn-herg
+" Ready dmpnn-herg Prediction of hERG Channel Blockers with Directed Message Passing Neural Networks ['Cardiotoxicity', 'hERG', 'Toxicity', 'Descriptor'] Single Probability of blocking hERG (cut-off: 10uM) ['Classification'] leilayesufu Single ['Float'] https://hub.docker.com/r/ersiliaos/eos30f3 ['AMD64'] https://ersilia-models-zipped.s3.eu-central-1.amazonaws.com/eos30f3.zip {'label': 'GitHub', 'url': 'https://github.com/ersilia-os/eos30f3'} "$ ersilia serve dmpnn-herg
 $ ersilia api -i 'CCCOCCC'
 $ ersilia close" 2023-12-04 https://github.com/leilayesufu
 eos80ch ['Compound'] Pretrained https://github.com/ersilia-os/eos80ch https://pubs.acs.org/doi/10.1021/acsomega.3c05664 https://github.com/M2PL GPL-3.0 ['Probability'] "Prediction of the antimalarial potential of small molecules using data from various chemical libraries that were
@@ -378,7 +382,7 @@ eos2rd8 ['Compound'] Pretrained https://github.com/ersilia-os/eos2rd8 https://ar
 $ ersilia api -i 'CCCOCCC'
 $ ersilia close" 2022-11-14 https://github.com/Amna-28
 eos7kpb ['Compound'] In-house https://github.com/ersilia-os/eos7kpb https://www.nature.com/articles/s41467-023-41512-2 https://github.com/ersilia-os/h3d-screening-cascade-models GPL-3.0 ['Probability'] "This panel of models provides predictions for the H3D virtual screening cascade. It leverages the Ersilia Compound Embedding and FLAML. The H3D virtual screening cascade contains models for Mycobacterium tuberculosis and Plasmodium falciparum IC50 predictions, as well as ADME, cytotoxicity and solubility assays
-" Ready h3d-virtual-screening-cascade-light H3D virtual screening cascade light ['Malaria', 'P.falciparum', 'Tuberculosis', 'M.tuberculosis', 'ADME', 'Cytotoxicity', 'Solubility'] Single The raw scores are the ones emerging from the FLAML model. The ones with a sufix _perc represent the percentile in the scale 0-1 over a ChEMBL dataset of 200k compounds. ['Classification'] miquelduranfrigola List ['Float'] {'label': 'GitHub', 'url': 'https://github.com/ersilia-os/eos7kpb'} "$ ersilia serve h3d-virtual-screening-cascade-light
+" In progress h3d-virtual-screening-cascade-light H3D virtual screening cascade light ['Malaria', 'P.falciparum', 'Tuberculosis', 'M.tuberculosis', 'ADME', 'Cytotoxicity', 'Solubility'] Single The raw scores are the ones emerging from the FLAML model. The ones with a sufix _perc represent the percentile in the scale 0-1 over a ChEMBL dataset of 200k compounds. ['Classification'] miquelduranfrigola List ['Float'] https://ersilia-models-zipped.s3.eu-central-1.amazonaws.com/eos7kpb.zip {'label': 'GitHub', 'url': 'https://github.com/ersilia-os/eos7kpb'} "$ ersilia serve h3d-virtual-screening-cascade-light
 $ ersilia api -i 'CCCOCCC'
 $ ersilia close" 2023-05-09 https://github.com/miquelduranfrigola
 eos2a9n ['Compound'] Online https://github.com/ersilia-os/eos2a9n https://www.frontiersin.org/articles/10.3389/fchem.2020.00046/full http://130.92.106.217:8080/chemblMuti.v1/ None ['Compound'] "Given a molecule, this model looks for its 100 nearest neighbors in the ChEMBL database, according to ECFP4 Tanimoto similarity. Due to size constraints, the model redirects queries to the ChEMBL server, so when using this model predictions are posted online.
@@ -593,7 +597,7 @@ eos157v ['Compound'] Pretrained https://github.com/ersilia-os/eos157v https://pa
 $ ersilia api -i 'CCCOCCC'
 $ ersilia close" 2022-07-13 https://github.com/Amna-28
 eos4qda ['Compound'] Pretrained https://github.com/ersilia-os/eos4qda https://jcheminf.biomedcentral.com/articles/10.1186/s13321-021-00566-4 https://github.com/UnixJunkie/FASMIFRA GPL-3.0 ['Compound'] "FasmiFra is a molecular generator based on (deep)SMILES fragments. The authors use Deep SMILES to ensure the generated molecules are syntactically valid, and by working on string operations they are able to obtain high performance (>340,000 molecule/s). Here, we use 100k compounds from ChEMBL to sample fragments. Only assembled molecules containing one of the fragments of the input molecule are retained.
-" Ready fasmifra FasmiFra molecule generator ['Compound generation'] Single 1000 generated molecules per each input ['Generative'] miquelduranfrigola List ['String'] https://hub.docker.com/r/ersiliaos/eos4qda ['AMD64', 'ARM64'] https://ersilia-models-zipped.s3.eu-central-1.amazonaws.com/eos4qda.zip {'label': 'GitHub', 'url': 'https://github.com/ersilia-os/eos4qda'} "$ ersilia serve fasmifra
+" Ready fasmifra FasmiFra molecule generator ['Compound generation'] Single 1000 generated molecules per each input ['Generative'] miquelduranfrigola List ['String'] https://hub.docker.com/r/ersiliaos/eos4qda ['AMD64'] https://ersilia-models-zipped.s3.eu-central-1.amazonaws.com/eos4qda.zip {'label': 'GitHub', 'url': 'https://github.com/ersilia-os/eos4qda'} "$ ersilia serve fasmifra
 $ ersilia api -i 'CCCOCCC'
 $ ersilia close" 2023-08-01 https://github.com/miquelduranfrigola
 eos2b6f ['Compound'] Pretrained https://github.com/ersilia-os/eos2b6f https://www.biorxiv.org/content/10.1101/2022.01.20.476787v1 https://github.com/mayrf/pkasolver MIT ['Experimental value'] "This model employs transfer learning with graph neural networks in order to predict micro-state pKa values of small molecules. The model enumerates the molecule's protonation states and predicts its pKa values. It was trained in two phases, first, using a large ChEMBL dataset and then fine-tuning the model for a small training set of molecules with available pKa values. The model in this repository is the pkasolver-light, which does not require an Epik license and is limited to monoprotic molecules.