I think it would depend heavily on what type of toxicity you are after...

Hello. We were thinking initally of LD50 calculation but this is a new area for our group.

In general, a LD50 value has little to no value... it's about the crudest form of toxicity existing and one cannot determine them in the EU

As you're talking about cell culture and MTT, I guess that you want to determine EC50 values for your coumpound. If so, you may use a software like sigmaplot to plot concentrations vs effect and then choose the appropriate non-linear regression model to derive your EC50 values

Sorry - haven't phrased this very clearly. We already calculate EC50's from the cells but what we would like to do is to try to model these compounds in a bioinformatics / software package to see if we can predict the toxicity and then compare to the EC50 data we already have with the view to using it to predict toxicity for other compounds we intend to synthesise.

Hello Caroline, you may find this useful, is a good stat software for EC50 estimations

http://www.biostat.uni-hannover.de/stat4tox.html

Hello Caroline,

My proper advice: you have not enough data to address some prediction toxicity software. To look only on EC50 values for cell toxicity is wasting of time! If you have more than 15 tested compounds you can try to do QSAR study, although the simple value of EC50 for cell toxicity is not suitable value (better enzyme inhibition parameters etc....

Stan

Caroline, do you mean you would like software that takes a chemical compound and predicts toxicity? If so, just do a search for "toxicity prediction" and you will see several prediction programs and web sites.

I've never used this kind of predictive modeling software, but it seems to be a very active area of research.

Some examples I know of:

http://www.opentox.org/toxicity-prediction

http://insilicofirst.com/

And a short list at http://www.predictive-toxicology.org/programs.html

As to which one(s) is better then another, I have no idea, and I'm not sure there even is an answer to that question yet. Also, some of these tools seem to have been designed in a manner that may make them better at some classes of chemicals than others (eg. some seem designed to predict drug toxicity, so may not do so well with say, a pesticide or environmental pollutant).

Hello Caroline,

I agree with Michael and Peter: everything depends on what kind of toxicity (cell, system or specific) are you looking for. As you've mentioned that LD50 calculation is new for you, let me share some additional information from my personal experience (I am testing natural phytochemicals and synthesized compounds both in vitro and in vivo for years): 1) very often the results obtained in vitro (in cell culttoures, isolated organs) don't correlate with those in vivo, therefore it's very difficult to predict the toxicity of a compound based on information obtained solely in in vivo or in vivo experiment; 2) values of LD50 will vary depending on i) route of administrtion (i.v., s.c., i.p., oral, topical, etc.); ii) animal (mice, rats, rabbits, guinea pigs); 3) LD50 as well as effective doses obtained in animal experiment can't be dirctly extrapolated to humans (e.g. http://www.fasebj.org/content/22/3/659.full). 4) Very important point: amount of the compound to be tested for LD50 - you shoud have at least 150 mg to perform the simpliest test on mice (minimal #of animals-12).

Concerning the software I would like recommend the following resources

http://www.pharmaexpert.ru/PASSOnline/ for rapid prediction of possible activity and

https://ilab.acdlabs.com/iLab2/ for more detailed data

Both are free, but ACD Labs has also very powerful commersial software.

If you have any additional questions, I'll be glad to help any way I can.

Best wishes

Hi All, Some very useful answers here! I agree that LD50 is probably not going to be the aim of your modelling experiments and that a more QSAR like approach is probably the way to go. Perhaps narrow down the toxicological outcome that is being predicted such as genotox or interaction with a target organ. Then input physicochemical data to one of the many modelling paradigms. (and cross fingers).

Thank you everyone for your advice, this will help us to get going. Caroline

Hi, DEREK is used widely to predict toxicity using QSAR approach. It is also approved by FDA..

Try this one:

http://www.opentox.org/dev/documentation/components/toxtree

it is perhaps the most comprehensive software tool for toxicity estimation

You may use many of online software, such as:

FAF-Drugs2 (http://mobyle.rpbs.univ-paris-diderot.fr/cgi-bin/portal.py?form=FAF-Drugs#forms::FAF-Drugs2) It could be used to predict toxicity and oral bioavailability.

LAZAR (http://LAZAR.im-silico.de) . You can use it to predict mutagenicity and carcinogenicity of the compounds.

ACD/I-Lab (https://ilab.acdlabs.com/). It is a very powerful program. It can predict variety of compound properties such as bioactivity, oral bioavailability, health effect probability, and maximum recommended daily dose.

Toxtree (http://toxtree.sourceforge.net) you can also predict the mutagenicity and carcinogenicity of the compounds using this software.

Hope this answer may help your research :D

Hi Caroline,

I work for a software company in India.We have a standalone software for calculating ecotoxicity, based on ETA indices. There is a demo version available here: http://aptsoftware.com/ChemStudio_QT_V1.0.3(D).zip

You can have a look. I can modify it according to your need, if you think this will serve your purpose.

We also have a rmsquare calculator site to calculate the quality of your predicted model: http://aptsoftware.co.in/rmsquare

Regards

Pratim Chakraborti

You can try "drc" package in "R", a wonderful open source package. There are also other packages like "grofit", "drfit" etc. I find drc handy and versatile.

Dear Colleagues, 

My question goes in the same direction as the one done by Caroline, but I think that the main problem might not the same... I am working with mixtures of unknown molecules characterised by UHPLC-HRMS and it is my willingness to study its toxic properties. I have already performed MICROTOX assays with vibrio fischeri (they are toxic as a mixture)and I was wondering the possibility to work with cell lines. However, what I would like to know if is there any soft capable to calculate for example EC50 or IC50 based on the chemical structure.

I have already it with the ACDlabs software and the rest posted here but results found does not convince me a lot... 

Thanks for your time

Javier

Compusyn is a software (free) based on the concept of combination index (CI) and its isobologram algorithm. This software can be used in the study of chemical mixtures. Please follow the link below:

http://www.combosyn.com/

Thanks Syamkumar!

osiris property explorer server may be useful to you. link http://www.organic-chemistry.org/prog/peo/

I think http://tox.charite.de/protox_II/ can be helpful for researchers who have plan to follow this mainstream line. Of course, some commercial tools, for example Discovery studio and Schrodinger package, can easily provide modules for the users to calculate the associated toxicity with natural or synthetic products. To me, the discovery studio is the best one for this propose the researchers have.

You may try Toxicity Estimation Software Tool (TEST):

https://www.epa.gov/chemical-research/toxicity-estimation-software-tool-test

good article for discussing prediction of toxicity of chemicals