| thresholds and weights of toxicity risk [message #1986] |
Tue, 22 August 2023 17:45  |
nbehrnd
Messages: 211
Registered: June 2019
|
Senior Member |
|
|
Dear Thomas,
the generation of evolutionary libraries can be restrained by fitness criterion «toxicity risk». Based on some preliminary tests, I have difficulties to comprehend the thresholds available. Departing from «nasty functions» as one of DW's calculated properties, one could assume DW counts functional groups. The suggested default is 2, an integer. Then, choosing a toxicity risk less or equal to zero would render the restraint a constraint.
Interestingly, however, DataWarrior equally permits to enter an upper threshold as a real/floating number (a negative one like `-1.23` does not hinder DW to work either).
Can you please share some insight how to engage well these thresholds? Do the weights act on a linear penalty scale to provide the fitness criterion greater influence on the overall fitness of the molecules suggested? (DW's internal documentation (via F1) includes the string «toxicity» only once.)
Best regards,
Norwid
[Updated on: Tue, 22 August 2023 17:48] Report message to a moderator |
|
|
|
| Re: thresholds and weights of toxicity risk [message #1990 is a reply to message #1986] |
Thu, 24 August 2023 14:58  |
thomas
Messages: 655
Registered: June 2014
|
Senior Member |
|
|
Dear Norwid,
I admit that fitness criteria are not covered by the manual, because originally they were very simple.
The calculated combined toxicity risk of a compound is simply the sum of risk contributions from the 4 toxicity classes (mutagenic, tumorigenic, irritant, and reproductive effective. A low risk contributes 1, a higjh risk contributes 2 to the combined score. Thus, a non-existing combined risk would be 0 and the very unlikely theoretical maximum is 8.
Threshold may be fractional numbers, because they describe the sigmoid curve that translate the risk (0...8) into a fitness (1...0)
The determination, whether a compound is assigned a no, low or high risk is based on a statistical approach based on whether and which fragments are present from a predefined fragment list, which are frequently found in toxic compound of the particular toxicity class.
Hope, this explains it somewhat...
[Updated on: Thu, 24 August 2023 14:58] Report message to a moderator |
|
|
|
|
|
Search
Help
Members
Register
Login
Home
