Utilities

Other descriptions coming soon

Solubility Prediction and Solvent Screening

ASKCOS implements the machine learning solubility estimator of Vermeire, Chung, and Green (ACS Publications). The solubility prediction utility allows for both individual and batch calculation of solid solubility, requiring just the SMILES representation of solute and solvent and the desired temperature. The solvent screening utility provides for easy screening of the solubility of a single solute in a range of solvents and at a range of temperatures. See the utility's "Model I/O Details" tab for more input and output details.

Brief model description

This model first estimates the solubility at 298 K of the input molecule in water and then corrects this solubility to a given solvent by using the difference of solvation free energies (\Delta G_{solv}) in water and the requested solvent. If the solubility of the solute is known in another solvent, this information can be supplied and used instead of the water solubility prediction. Temperature dependence is calculated using dissolution enthalpy (the enthalpy change when dissolving a solute from the solid to solvent), which in turn is calculated via a thermodynamic cycle using ML prediction of sublimation enthalpy, solvation enthalpy, and solid and gas phase heat capacities. If known, experimental values for these heat capacities and sublimation enthalpy can be used as inputs to the model for more accurate prediction. Further details can be found in the referenced publication.

Limitations

The model most accurately solubility trends. In cases where the absolute solubility prediction is off, the relative solubility prediction between solvents is usually still good. Predictions are limited to neutral solvents in the liquid phase and neutral solutes in the solid phase. Temperature dependence above moderate temperatures (~350K) is only available for ~100 solvents whose critical temperatures and critical densities are known.

Buyable Compounds

Upload new buyables/building blocks

ASKCOS only support JSON /uploads of the buyables. NOTE: Only the admin user(s) have the ability to add buyables. Below are the necessary keys required to create a successful upload. Additional properties can be added into the "properties" key, eg. links, availability etc.

[
    {
        "smiles": "CCC",
        "ppg": 1,
        "source": "test source 1",
        "lead_time": "1day",
        "properties": [
            {"link": "https://test.com/query=CCC"},
            {"availability": "In-stock"}
        ]
    },
    {
        "smiles": "CCC",
        "ppg": 1.5,
        "source": "test source 2",
        "lead_time": "1day",
        "properties": [
            {"link": "https://test1.com/query=CCC"},
            {"availability": "In-stock"}
        ]
    }
]

Buyables/Building Block Description

Buyables data are either extracted from the website or through third party collaboration.

Shorthand Notation	Actual Name	Last Updated (Version)	Count
EM	eMolecules	2017 (ASKCOSv1)	110k
SA	Sigma Aldrich	2017 (ASKCOSv1)	20k
LN	LabNetwork	2017 (ASKCOSv1)	152k
MC	Mcule	Oct 2023	410k
CB	ChemBridge	Oct 2023	17k

MongoDB _id creation

The _id field created for each of the building blocks in those molecules is done using this code in python.

import hashlib
smi_vendor = f"{smiles}{VENDOR}"
hash_id = hashlib.sha256(smi_vendor.encode('utf-8')).hexdigest()