EstiMol - GWP, CED and Eco-Indicator99 of chemicals based on molecular structure
EstiMol is a free database from ifu Hamburg for Estimating LCI(A) data from the Molecular structure of chemicals. EstiMol allows to fill data gaps caused by poor data availability of LCIs for chemicals without requiring knowledge on the production process.
EstiMol features cradle-to-gate Global Warming Potential (GWP), Cumulated Energy Demand (CED) and EcoIndicator 99 for roughly 14.000 substances. These indicators were calculated using artificial neural networks.
Check out EstiMol
The use of EstiMol is currently free of charge. Just contact us to get the link. After a short registration you can search the database for (trivial) names, CAS registry numbers and even the structural formula: Just draw a molecule and search for it directly. Do you wish to use EstiMol free of charge?
Where/how can EstiMol data be used?
EstiMol supplies values for GWP, CED and Ecoindicator99. You can use these data e.g. in a Carbon Footprint project or a CED study for chemicals that you expect to have a minor impact on the overall results. When using the data, please cite Wernet et al. 2009  as reference for the calculation method.
The calculated data can also be used for priority setting: When several data gaps are present due to missing LCI data, the calculated values can be compared to support decision making regarding choosing substances to examine further (as presented in ).
Can indicators be calculated for all existing chemicals?
The underlying methodology developed in the FineChem  tool on which EstiMol is based has of course constraints. Artificial Neural Networks are very powerful when it comes to determining fuzzy correlations. Their weak point is extrapolation. Hence, the range of validity strongly depends on the range of chemicals that were used to train the neural networks.
The tool is only valid for petrochemicals that fulfill a set of preconditions: No bromine or iodine, not more than four fluorine atoms and not more than one sulphur or phosphorous atom may be contained in the molecule.
Furthermore the calculated results should be within a range that was present in the training datasets. As the neural networks were trained separately for GWP, CED and EcoIndicator 99, it is possible that one of the values is outside this range while another is valid. One value outside the range of validity can be a hint that the results are less reliable. It is therefore not recommended to use them. While using EstiMol, information about results exceeding the range of validity as well as other reasons causing the compound to be not suitable for the method will be displayed on the results page.
What are neural networks and how are they used for EstiMol?
The FineChem Tool  uses neural networks to calculate GWP, CED and EcoIndicator99 values. Basically, artificial neural networks (leaning towards the concept of biological neural networks) consist of many small units which in sum are able to solve complex problems. This statistical method is especially suitable when a correlation is present but there is no known cause-effect mechanism.
This is the case for LCIA results of chemicals and their molecular structure. An algorithm was applied to train the neural networks by adjusting parameters of the networks. Results of about 330 high-quality LCAs were used for the training.
Further information on EstiMol
What do you think about EstiMol? Do you have questions regarding its use? Do you need further information, or are you interested in further developing EstiMol to cover additional chemical compounds? Then please contact us. We are looking forward to speaking with you.
Funding, data sources and underlying work
EstiMol was developed within the project 'Strukturbasierte Umweltbewertung von Chemikalien' (DBU-AZ 30182-31). We would like to thank Deutsche Bundesstiftung Umwelt (DBU) for their support.
The GWP, CED and Ecoindicator 99 values contained in EstiMol were calculated with the FineChem-Tool . This tool was developed by Gregor Wernet at ETH Zurich (chair Prof. Hungerbühler) and took a unique approach on estimating LCI(A) indicators by using artificial neural networks.
A detailed description of the methods used can be found in . We are grateful, that Gregor Wernet provided the methods and his knowledge for EstiMol.
Chemicals contained in EstiMol and their names and CAS numbers were taken over from the database PubChem .
 Safety and Environmental Technology Group (2010): The Finechem Tool. www.sust-chem.ethz.ch/tools/finechem
 G. Wernet, S. Papadokonstantakis, S. Hellweg, K. Hungerbühler (2009): Bridging data gaps in environmental assessments: Modeling impacts of fine and basic chemical production, Green Chem, 11 (11), 1826.
 National Center for Biotechnology Information. The PubChem Project. pubchem.ncbi.nlm.nih.gov.
 G. Wernet, S. Hellweg, K. Hungerbühler (2012): A tiered approach to estimate inventory data and impacts of chemical products and mixtures. Int J LCA 2012;17(6):720–8