IN-VITRO SUSCEPTIBILITY TESTING AND QUANT ITATIVE STRUCTURE-ACTIVITY-RELATIONSHIPS (QSAR) FOR THE DETERMINATION OF ANTIASPERGILLUS ACTIVITYFOR ANALOGS OF GLYCOLIPIDS
Jc. Garrigues et al., IN-VITRO SUSCEPTIBILITY TESTING AND QUANT ITATIVE STRUCTURE-ACTIVITY-RELATIONSHIPS (QSAR) FOR THE DETERMINATION OF ANTIASPERGILLUS ACTIVITYFOR ANALOGS OF GLYCOLIPIDS, Journal de mycologie medicale, 6(3), 1996, pp. 111-117
Objective. A quantitative structure activity relationship by neural ne
twork was used to dispose the activity of new active molecules against
Aspergillus fumigatus. A structure-activity relationship try to conne
ct molecular properties represented by different parameters to antiasp
ergillus activity. Lop P (lipophilicity parameter) and Rz/c (ratio bet
ween number of heteroatoms and carbon atoms) are essential parameters,
associated to connectivity or topological parameters, showing molecul
ar length, ramification number, cycle number. Materials and methods. T
he antifungal activity, represented by the concentration which reduce
by 90 % the growth of A. fumigatus (IC90), is measured with an in vitr
o test, based on glucose consumption. To make the relation between the
structure and the antiaspergillus activity, we used 2 commercial comp
ounds: amphotericin B and itraconazole with 13 amphiphilic glycolipid
analogs. The polar head derive from glucose and lactose, the hydrocarb
on chain also varying. We connected molecular parameters and antifunga
l activity with a computer assisted system, based on artificial intell
igence: a neural network. When the structure-activity relationship is
done, with experimental results (IC90 measured with an in vitro test),
we calculated predictive values for new antiaspergillus amphiphilic c
ompounds, with molecular modeling, for the synthesis of interesting st
ructures. Results. Absolute value of the difference between experiment
al IC90 (measured with an in vitro test), and a predicted IC90 with a
neural network, varies from 1.74 to 6.88 mu mol./l. Conclusion. The in
terest in the use of neural network to predict antifungal activity of
sugar derived amphiphilic compounds, against A. fumigatus was clearly
shown in this study. With a neural network, it is possible to predict
an IC90, before to synthesize a new molecule. The calculated IC90 are
in agreement with the IC90 measured with an in vitro test.