Model incorporation workflow

Example of an end-to-end model incorporation workflow

Overview

The model incorporation pipeline is streamlined by a series of GitHub Actions that will automatize model testing and updating of our backend services (Airtable, DockerHub and AWS), simplifying the process for model contributors. The different workflows involved can be found in:

Reusable workflows: the actual workflows are stored in a single repository, ersilia-model-workflows, so any change done in those will easily propagate to all models.
Caller workfflows: the reusable workflows are called by the workflows in eos-template upon Pull Request (PR) or Push to the model repository.
Model request workflow: the model incorporation pipeline is triggered when a Model request issue is approved by an Ersilia maintainer. This workflow is part of Ersilia's main repository workflows.

More details on the actual workflows can be found in the Developers section. Here we are just giving an overview of the steps a model contributor needs to take to successfully include a new model in Ersilia.

A toy example

If you want to test the workflow with a toy example, we have prepared an Ersilia Demo Repository. This repository is focused on illustrating the GitHub Actions workflows to be followed in order to successfully incorporate a model to the Ersilia Model Hub.

1. Create a model request at Ersilia

Go to the Ersilia main repository issues page.
Click on New issue. Then 🦠 Model Request (Get started)
For the purpose of this demo, you can use the following information:
- 🦠 Model Request: Demo Malaria Model
- Model Name: Demo Malaria Model
- Model Description: Prediction of the antimalarial potential of small molecules. This model was originally trained on proprietary data from various sources, up to a total of >7M compounds. The training sets belong to Evotec, Johns Hopkins, MRCT, MMV - St. Jude, AZ, GSK, and St. Jude Vendor Library. In this implementation, we have used a teacher-student approach to train a surrogate model based on ChEMBL data (2M molecules) to provide a lite downloadable version of the original MAIP.
- Slug: demo-malaria-model
- Tag: Malaria,P.falciparum
- Publication: https://jcheminf.biomedcentral.com/articles/10.1186/s13321-021-00487-2
- Code: https://www.ebi.ac.uk/chembl/maip/
- License: GPL-3.0-only

2. Wait until model approval

The Ersilia team will revise your model requests and likely start a public discussion around it.
At some point, your model will be approved. You will see an /approve comment in the GitHub issues.
Approval will trigger some GitHub Actions. Eventually, the ersilia-bot will post an informative message in your issue. Importantly, this message will contain a link to a new model repository placeholder, named, for example, ersilia-os/eosXabc. This code is arbitrarily assigned by Ersilia and will change every time. You should not worry about creating one manually and you should never modify it.

3. Fork the model repository

Go to the model repository page: https://github.com/ersilia-os/eosXabc, in this case.
Fork the repository to your GitHub user.
Clone the forked repository. This should create a eosXabc folder in your local filesystem.

4. Use this demo model

For this demo, you have to clone the current repository (ersilia-os/eos-demo). This will create an eos-demo folder in your local filesystem.
Run the following script to populate your forked model with the demo data: python /path/to/eos-demo/populate.py /path/to/eosXabc.
Your eosXabc has been populated with model parameters, dependencies and extra metadata. It is now ready for commit.

5. Make a pull request to Ersilia

Commit changes and push changes to eosXabc.
Open a PR to the main branch at ersilia-os/eosXabc. GitHub Actions workflows will be triggered to ensure that your code works as expected.

6. Wait until model is merged

The Ersilia team will revise your PR and merge it eventually. More GitHub Actions workflows will be triggered at this point.
Once the model is merged, you should see it in Ersilia's AirTable.

7. Assist with curation and publication

The ersilia-bot will open a new issue at ersilia-os/eosXabc. As you will see, someone from the Ersilia community will be assigned as a reviewer of the model.
If you are a member of the Ersilia Slack workspace, then you may also see activity triggered around your model.

A real-world example

Now that we have a general idea of the contents of the Ersilia Model Template, we will follow the example of a simple but widely used model to calculate the synthetic accessibility of small molecule compounds. Synthetic accessibility measures the feasibility of synthesizing a molecule in the laboratory. In 2009, Peter Ertl presented the synthetic accessiblity (SA) score, based on measures of molecular complexity and occurrence of certain fragments in the small molecule structure. High (greater than 6) SA scores denote difficult-to-synthesize molecules, and low (lower than 3) SA scores suggest that the molecule will be easy to synthesize.

1. Open a Model Request Issue

Please fill in the issue fields as accurately as possible and wait for review and approval by one of the Ersilia maintainers.

Read the publication

It is important that you read the original publication in order to understand the training data, the limitations of the model, the validation performed and the characteristic of the algorithm, among other details. In this case, this is a classic (old) publication from a Novartis team. They analyzed small fragments in the PubChem database and devised a molecular complexity score that takes into account molecule size, presence of non-standard structural features, presence of large rings, etc. The authors validated the model by comparing their SA score with synthetic feasibility as estimated by medicinal chemists for a set of 40 molecules.

Find model code and parameters

Code to calculate the SA score does not seem to be available from the publication. Fortunately, though, the RDKit library, in its contributions module, contains an implementation of the SA score. The code can be found here. This RDKit-based implementation was developed in 2013 by Peter Ertl and Greg Laundrum.

2. Run the code outside Ersilia

Before incorporating the sa-score model to the Ersilia Model Hub, we need to make sure that we can actually run the code provided by the third party. In this case, upon quick inspection, two elements seem to be central in the repository:

The sascorer.py script, containing the main code. We can consider this file to be the model code.
The fpscores.pkl.gz compressed file, containing pre-calculated fragment scores. In this simple case, we can consider this file to be the model parameters.

Create a conda environment

No installation instructions are provided for this model. However, the sascorer.py file import statements indicate that, at least, rdkit is necessary. We can create a Conda environment and install the rdkit library as follows:

conda create -n sa-score python=3.10
conda activate sa-score
pip install rdkit

At this stage it is useful to note down the specific versions of each package required, as Ersilia requires them.

Download code and parameters

We can download code and parameters of the directly from the RDKit contributions repository. Here, we store these data in a folder named SA_Score located in the ~/Desktop.

cd ~/Desktop
mkdir SA_Score
cd SA_Score
wget https://raw.githubusercontent.com/rdkit/rdkit/master/Contrib/SA_Score/sascorer.py
wget https://raw.githubusercontent.com/rdkit/rdkit/master/Contrib/SA_Score/fpscores.pkl.gz

Often, the model will be available as a full GitHub repository. In these cases, you can simply clone the repository.

Test the model

Inspection of the sascorer.py file indicates that we can run this script from the terminal:

sascorer.py

...
if __name__ == '__main__':
    import sys
    import time

    t1 = time.time()
    readFragmentScores("fpscores")
    t2 = time.time()

    suppl = Chem.SmilesMolSupplier(sys.argv[1])
    t3 = time.time()
    processMols(suppl)
    t4 = time.time()

    print('Reading took %.2f seconds. Calculating took %.2f seconds' % ((t2 - t1), (t4 - t3)),
          file=sys.stderr)

The Chem.SmilesMolSupplier takes a file containing SMILES strings and identifiers, separated by a space character. The file expects a header. Let's create a file with a few molecules. We looked for examples in DrugBank:

molecules.smi

smiles identifier
CC(=O)NC1=CC=C(O)C=C1 mol1
CN(C)CCC1=CNC2=C1C=C(CS(=O)(=O)N1CCCC1)C=C2 mol2
CCN(CC)CC1=C(O)C=CC(NC2=C3C=CC(Cl)=CC3=NC=C2)=C1 mol3

Now let's see if the model works as expected:

python sascorer.py molecules.smi

It does work! The output in the terminal looks like this:

smiles	Name	sa_score
CC(=O)Nc1ccc(O)cc1	mol1	1.407299
CN(C)CCc1c[nH]c2ccc(CS(=O)(=O)N3CCCC3)cc12	mol2	2.319770
CCN(CC)Cc1cc(Nc2ccnc3cc(Cl)ccc23)ccc1O	mol3	2.249844
Reading took 0.21 seconds. Calculating took 0.00 seconds

Many repositories give a clear description of the expected input format. For the SA scorer, the expected input was not clearly specified, and previous knowledge of the Chem.SmilesMolSupplier method was necessary.

3. Fork the new model repository based on the Ersilia Model Template

Now that we know that the code can run in our local machine, we can fork the new repository created by the Model Request Workflow and start working on it.

Fork and Clone the new repository

Always fork the repository to your user, and clone it from there in our local machine. Let's do it in the ~/Desktop:

cd ~/Desktop
git clone https://github.com/user-github/eos9ei3.git
cd eos9ei3

Migrate code and parameters

Let's now place the code and the parameters in the model folder (in the framework and checkpoints sub folders, respectively):

cd ~/Desktop
cp SA_Score/sascorer.py eos9ei3/model/framework/code/.
cp SA_Score/fpscores.pkl.gz eos9ei3/model/checkpoints/.

The github repository contains a dummy mock.csv that can be deleted. Also remove the .gitattributes file if it is not needed.

Note that here we are migrating code and parameters to different folders. This may cause critical errors if code expects to find parameters at a certain relative location. Try to locate the pointers to the model parameters and change the paths. Only, and exceptionally, if the model architecture is too complex we can keep code and parameters in the /framework folder. Please ask for permission to Ersilia's team before doing it.

Write framework code

Now it is time to write some code. Here we will follow the description of the model folder given in the Model Template page.

The eos-template provides a main.py that will guide us through the deployment steps. We will adapt the main.py to our specific needs, and run it from the run.sh file present in the /model directory. The arguments are, respectively, the path to the file, the input and the output. Ersilia takes care of passing the right arguments to the run.sh file. Please do not modify it.

python $1/code/main.py $2 $3

Going back to our model of interest, we have identified three necessary steps to run the model:

Input adapter
SAScorer (in this case, the sascorer.py which is already written)
Output adapter

Write the input adapter

By default, for chemical compound inputs, Ersilia uses single-column files with a header. However, the sascorer.py expects a two-column file. Let's write an input adapter. The best practice to create temporary files is to include a temporal folder with the prefix "ersilia" and remove it at the end of the model run.

input_adapter.py

import sys
import csv
import tempfile

input_file = sys.argv[1] 

root = os.path.dirname(os.path.abspath(__file__))
temp_dir = tempfile.mkdtemp(prefix="ersilia_", dir=root)
temp_input = os.path.join(temp_dir, "tmp_input.smi")

smiles_list = []
with open(input_file, "r") as f:
    reader = csv.reader(f)
    next(reader) # skip header
    for r in reader:
        smiles_list += [r[0]]

with open(temp_input, "w") as f:
    writer = csv.writer(f, delimiter=" ")
    writer.writerow(["smiles", "identifier"]) # header
    for i, smi in enumerate(smiles_list):
        writer.writerow([smi, "mol{0}".format(i)])

The script creates an intermediate tmp_input.smi file that can be used as input for sascorer.py. We can keep this as a separate script under /code, but since it is a small function, we will write it inside the main.py file:

code/main.py from line 20

# read SMILES from .csv file, assuming one column with header
smiles_list = []
with open(input_file, "r") as f:
    reader = csv.reader(f)
    next(reader) # skip header
    for r in reader:
        smiles_list += [r[0]]

with open(temp_input, "w") as f:
    writer = csv.writer(f, delimiter=" ")
    writer.writerow(["smiles", "identifier"]) # header
    for i, smi in enumerate(smiles_list):
        writer.writerow([smi, "mol{0}".format(i)])

Make sure that parameters are read

So far, we haven't pointed to the model parameters. When migrating code and parameters, we separated the sascore.py file and the fpscores.pkl.gz file.

Let's inspect sascore.py to understand how parameters are read. There is a readFragmentScore function that does this job. We need to modify it to point to the checkpoints folder:

code/sascorer.py

...
def readFragmentScores(name='fpscores'):
    import gzip
    global _fscores
    # generate the full path filename:
    if name == "fpscores":
        name = op.join(op.join(op.dirname(__file__), "..", "..", "checkpoints"), name)
    data = pickle.load(gzip.open('%s.pkl.gz' % name))
    outDict = {}
    for i in data:
        for j in range(1, len(i)):
            outDict[i[j]] = float(i[0])
    _fscores = outDict
...

Run the model inside main.py

We now have the input adapter and the model code and parameters. We simply need to run the model in main.py by calling the sascorer.py.

We first make sure to import the necessary packages and delete the non-necessary ones (in this case, the MW by RDKit). The sa-scorer uses time to measure how long did the model take, as well as the functions readFragmentScores and processMols.

The input file is no longer passed as a sys.argv, so we modify it with the temporal input file we just created

code/main.py

# imports
import os
import csv
import sys
import time
from rdkit import Chem
from sascorer import readFragmentScores, processMols

# parse arguments
input_file = sys.argv[1]
output_file = sys.argv[2]

# current file directory
root = os.path.dirname(os.path.abspath(__file__))
temp_dir = tempfile.mkdtemp(prefix="ersilia_", dir=root)
temp_input = os.path.join(temp_dir, "tmp_input.smi")

# read SMILES from .csv file, assuming one column with header
smiles_list = []
with open(input_file, "r") as f:
    reader = csv.reader(f)
    next(reader) # skip header
    for r in reader:
        smiles_list += [r[0]]

with open(temp_input, "w") as f:
    writer = csv.writer(f, delimiter=" ")
    writer.writerow(["smiles", "identifier"]) # header
    for i, smi in enumerate(smiles_list):
        writer.writerow([smi, "mol{0}".format(i)])

# run model
t1 = time.time()
readFragmentScores("fpscores")
t2 = time.time()

suppl = Chem.SmilesMolSupplier(temp_input)
t3 = time.time()
R = processMols(suppl)
t4 = time.time()

print('Reading took %.2f seconds. Calculating took %.2f seconds' % ((t2 - t1), (t4 - t3)),
        file=sys.stderr)

The processMols() function was simply printing the output without saving the results, so we have modified it in the sascorer.py to get the output:

def processMols(mols):
    print('smiles\tName\tsa_score')
    R = []
    for i, m in enumerate(mols):
        if m is None:
            continue
        s = calculateScore(m)
        smiles = Chem.MolToSmiles(m)
        print(smiles + "\t" + m.GetProp('_Name') + "\t%3f" % s)
        R += [[smiles, m.GetProp("_Name"), "\t%3f" % s]]
    return R

Write the output adapters

We need to understand the output of the model in order to collect it correctly. The easiest will be to add a print(R) statement in main.py and run the run.sh file with the example test file.

cd model/framework
python run.sh . examples/run_input.csv examples/run_output.csv

We observe that the output provided by sascorer.py has three columns (tab-separated), so we need to adapt it.

smiles

Name

sa_score

C(F)Oc1ccc(-c2nnc3cncc(Oc4ccc5ccsc5c4)n23)cc1

mol0

2.823995

C(F)Oc1ccc(-c2nnc3cncc(OCC[C]4BBBBBBBBBB[CH]4)n23)cc1

mol1

5.757383

Cn1cc2ccc(Oc3cncc4nnc(-c5ccc(OC(F)F)cc5)n34)cc2n1

mol2

2.910502

Likewise, we can substitute this piece of code in main.py:

code/main.py template line29

outputs = []
for r in R:
    outputs += [r[-1].strip()]
        
#check input and output have the same lenght
input_len = len(smiles_list)
output_len = len(outputs)
assert input_len == output_len

# write output in a .csv file
with open(output_file, "w") as f:
    writer = csv.writer(f)
    writer.writerow(["sa_score"])  # header
    for o in outputs:
        writer.writerow([o])

Finally, let's add one more line to main.py to clean up the temporary files we have created:

shutil.rmtree(temp_dir)

4. Run the model locally

Let's now check that the scripts run as expected. Eventually, Ersilia will run this code from an arbitrary location, so it is best to test it outside the framework folder. We can create an input file in the ~/Desktop.

molecules.csv

smiles
[H][C@]12SC(C)(C)[C@@H](N1C(=O)[C@H]2NC(=O)[C@H](N)C1=CC=CC=C1)C(O)=O
NC1=CC(=CNC1=O)C1=CC=NC=C1
ClC1=CC=C2N=C3NC(=O)CN3CC2=C1Cl
COC1=CC=C(C=C1)C1=CC(=S)SS1
COC1=CC=C(C=C1)C(=O)CC(=O)C1=CC=C(C=C1)C(C)(C)C

To test the model, we simply have to execute the run.sh script.

cd ~/Desktop
FRAMEWORK_PATH="eos9ei3/model/framework/"
bash $FRAMEWORK_PATH/run.sh $FRAMEWORK_PATH molecules.csv output.csv

You should get the output.csv file in your ~/Desktop. The output contains five predictions, corresponding to the five molecules in the molecules.csv file.

output.csv

sa_score
3.527006
2.344611
2.957539
2.535359
1.798579

Write the example files

If you haven't yet, make sure the run_input.csv (3 inputs) and run_output.csv (outcome of the bash run.sh command) are in model/framework/examples

Write the columns file

Modify the model/framework/columns/run_columns.csv file to include the information for each of the output columns of the model, in this case it would simply be:

name,type,direction,description
sa_score,float,low,Synthetic accessibility score

Edit the `install.yml` file

The install.yml file should include all the installation steps that you run after creating the working Conda environment. In the case of sa-scorer, we only installed RDKit. It is important to specify the right version of each package to avoid clashes.

install.yml

python: "3.10"

commands:
  - ["pip", "rdkit", "2023.3.1"]

Write the `metadata.yml` file

Don't forget to document the model. Read the instructions to write the metadata file page. The metadata.yml for this model should read like:

Identifier: eos9ei3
Slug: sa-score
Status: Ready
Title: Synthetic accessibility score
Description: "Estimation of synthetic accessibility score (SAScore) of drug-like molecules based on molecular complexity and fragment contributions. The fragment contributions are based on a 1M sample from PubChem and the molecular complexity is based on the presence/absence of non-standard structural features. It has been validated comparing the SAScore and the estimates of medicinal chemist experts for 40 molecules (r2 = 0.89). The SAScore has been contributed to the RDKit Package.\n"
Deployment: 
    - Local
Source: Local
Source Type: External
Task: Annotation
Subtask: Property prediction or calculation
Input:
    - Compound
Input Dimension: 1
Output:
    - Score
Output Dimension: 1
Output Consistency: Fixed
Interpretation: Low scores indicate higher synthetic accessibility
Tag:
    - Synthetic accessibility
    - Chemical synthesis
Biomedical Area:
    - Any
Target Organism:
    - Not applicable
Publication Type: Peer reiewed
Publication Year: 2021
Publication: https://jcheminf.biomedcentral.com/articles/10.1186/1758-2946-1-8
Source Code: https://github.com/rdkit/rdkit/tree/master/Contrib/SA_Score
License: BSD-3-Clause

5. Run the local model inside Ersilia

Before committing our new model to Ersilia, we must check it will work within the Ersilia environment. To do so, we have a very convenient option at model fetch time, --from_dir that allows us to specify a local path to the model we are fetching (so, instead of looking for it on GitHub it will use the local folder). It is crucial to complete this step before committing the model to GitHub.

conda activate ersilia
ersilia -v fetch eos9ei3 --from_dir ~/Desktop/eos9ei3
ersilia serve eos9ei3
ersilia predict -i molecules.csv -o output.csv

If this runs without issues, the model is ready to be incorporated. If not, please go to the Troubleshooting section to know what to do next.

6. Test the model

In addition to locally running the model, we advise going an step further and testing the model using the convenient test command. This will revise that the metadata is correct and all files necessary are present:

conda activate ersilia
ersilia -v test eos9ei3 --shallow --from_dir ~/Desktop/eos9ei3

This will create a report file stating if any test has not passed. To learn more about the different tests available, have a look at the Developers documentation.

7. Commit changes to the repository

We are now ready to commit changes, first to our fork, and then to the main Ersilia repository by opening a pull request. Before doing so, complete the steps below:

Cleanup mock files

Probably, there is a few files, such as mock.csv, that are no longer needed (this is used solely to initialize Git-LFS). Please remove them before committing and also eliminate the Git-LFS tracking from .gitattributes (see below for more on Git-LFS)

Check the `.gitattributes` file

If you have large files, you will have to track them with Git LFS. The template already provides a collection of common extensions to track with Git LFS (see .gitattributes file). However, it is possible that your parameters do not have any of these extensions.

Let's track our parameters, even if the file is not particularly big:

cd ~/Desktop/eos9ei3
git lfs track "*.pkl.gz"
git add .gitattributes
git commit -m "git lfs track"
git push

Now commit the bulk of your work:

git add .
git commit -m "first major commit"
git push

Open a pull request

Once the model is ready, open a pull request to merge your changes back into the main repository. This will trigger a series of GitHub Actions:

Security workflow: makes sure that no private key is released with the model
Json syntax check: controls that the metadata file does not contain Json syntax errors (does not check the content, only syntax)
Model Test on PR: this workflow will first test that the metadata.yml has the correct fields (if it doesn't, it will fail. Please look at the Action Run to get more information on why it has failed). If the metadata is correct, it will then go onto installing Ersilia and testing the model, similar to what we have done locally

If the Actions at Pull request fail, please check them and work on debugging them before making a new pull request. Ersilia maintainers will only merge PR's that have passed all the checks. Once the PR has the three green checks, the PR will be merged. This triggers a Model Test on Push action that will:

Test the model once more
Update the README.md and AirTable metadata

This workflow is also triggered each time there is a push to the repository, to ensure changes to the code do maintain model functionality and metadata is not outdated. If the model testing works, the model will then be zipped and uploaded to S3 as a backup and packaged in a Docker image and made available via the Ersilia DockerHub page. The Docker image is also tested extensively in the workflows before being added to the DockerHub.

You can now visit the eos9ei3 GitHub repository and check that your work is publicly available.

8. Fetch and serve the model with Ersilia

We are ready to test the model in the context of the Ersilia CLI. To run the model, simply run (using an example .csv file in the right format)

ersilia fetch eos9ei3
ersilia serve eos9ei3
ersilia run -i input.csv -o output.csv

The output should look like this:

 "key","input","sa-score"
 "RYYVLZVUVIJVGH-UHFFFAOYSA-N","Cn1cnc2n(C)c(=O)n(C)c(=O)c12",2.297982

Debugging the fetch and command of Ersilia can be very complicated. Read the Troubleshooting models page and, if you are still stuck, please open an issue in the model repository.

The original model contributor should make sure the model is working for different users and answer any questions or issues that might arise during model testing. If amends must be made, the original model contributor should work on those.

9. Clean up

Please, help us keep a healthy environment and avoid duplication of files and consuming of our Git LFS quota. Delete the repository fork after the model has been successfully incorporated.

PreviousModel template NextTroubleshooting models

Last updated 8 days ago

Was this helpful?

Overview

A toy example

1. Create a model request at Ersilia

2. Wait until model approval

3. Fork the model repository

4. Use this demo model

5. Make a pull request to Ersilia

6. Wait until model is merged

7. Assist with curation and publication

A real-world example

1. Open a Model Request Issue

Read the publication

Find model code and parameters

2. Run the code outside Ersilia

Create a conda environment

Download code and parameters

Test the model

3. Fork the new model repository based on the Ersilia Model Template

Fork and Clone the new repository

Migrate code and parameters

Write framework code

Write the output adapters

4. Run the model locally

Write the example files

Write the columns file

Edit the install.yml file

Write the metadata.yml file

5. Run the local model inside Ersilia

6. Test the model

7. Commit changes to the repository

Cleanup mock files

Check the .gitattributes file

Open a pull request

8. Fetch and serve the model with Ersilia

9. Clean up

Edit the `install.yml` file

Write the `metadata.yml` file

Check the `.gitattributes` file