QSPR prediction of pharmaceutical removal with GAC

1
QSPR prediction of pharmaceutical removal with GAC
David de Ridder
Carbamazepine
Delft
2
Problem statementmicropollutants

• 100.000 different micropollutants in surface
  water
• Variable mixture
• RIWA Rijn measures about 250 org. micropollutants
• Analysis programs costly and time consuming
• QSAR can be used as a screening tool

3
QSAR werkDavid de Ridder

• Singular processes
• Activated carbon (AC)
• Aeration
• Ozone
• UV
• Combinations of processes
• AC-UV/H2O2-AC
• (RBF)-NF-AC

4
Activated carbon Research objectives

• Estimate
• Influence of turbulence on adsorption kinetics
• Influence of water quality
• Influence of preloading carbon
• Construction of QSPR model
• Accuracy of QSPR prediction Ce, qe

5
Experiment set-uppharmaceutical selection
- 0
6
Experiment set-upmatrix
7
Experiment set-up

• 2 litre solution
• Kinetic _at_ 200 mg carbon
• Adsorption time 1 day 6
• weeks
• Equilibrium _at_ 20-2000 mg
• carbon
• Adsorption time 8 weeks
• Carbon 0,63-0,71 mm

8
ResultsKinetic experiments
9
ResultsKinetic experiments
10
ResultsKinetic experiments
11
ResultsKinetic experiments

• Lower turbulence decreases adsorption
  significantly
• Preloading large influence on negatives, and no
  significant effect on positives.

12
Results - Equilibrium
13
Results - Equilibrium
14
Results - Equilibrium
15
Resultsinitial remarks

• MW in range 200-300 D no significant influence
• Log D has higher impact on removal of negatives
• At similar pKa, higher log D yields higher
  removal
• At similar log D, positives are removed 1,2-2
  times more effective than negatives
• In wastewater
• Positives comparable removal as surface water
• Neutrals better removal than in surface water
• Significant removal of 4 negatives in blank!

16
Resultshypotheses

• Preloading creates a negatively charged layer
  onto the carbon, rejecting negatives and
  attracting positives
• In the MW range of 200-300 D, probably most
  carbon micropores will be available for
  adsorption
• In wastewater, (bio)degradation of negatives is
  preferred.

17
Results - QSPR construction

• 4 (out of 21) compounds excluded for verification
• MLR (multivariable linear regression) prediction
  model

18
Results - QSPR prediction
19
Results - QSPR validation
20
Results model accuracy
Consequent over/underprediction Less data
available
Specific mispredictions
21
Results Freundlich parameters
22
Initial conclusionsModel prediction

• Applied carbon dose too high -gt ultrapure models
  inaccurate
• Initial degradation negatives wastewater -gt
  negatives not taken properly into account
• Surface water general underprediction Ce at
  higher carbon dose
• Wastewater Specific overprediction Ce
  (Terbutaline, Salbutamol, metropolol,
  Clenbuterol, Aminopyrine)
• Demiwater PL consequent over/underprediction

23
Further research

• Dataset
• Larger variation MW (if relevant)
• Process conditions
• Carbon type
• Change preparation preloaded carbon
• Lower carbon doses
• pH variation at same water quality (decrease
  charge negatives, lower log D negatives higher
  log D positives)
• Analysis
• ATP to check for biological activity
• NOM characterisation (blank adsorbed (?))
• Carbon characterisation (PSD, hydrophobicity,
  carbon pKa)

24
Questions/remarks?
25
(No Transcript)
26
(No Transcript)
27
(No Transcript)
28
(No Transcript)