Credit is due to Dr Ed Delaney (Reaction Science Consulting LLC) and his co-workers from an array of top pharmaceutical companies for their recent publications on sulfonate ester formation.
From the outset the team, which was supported by PQRI, was determined to develop a mechanistic understanding to underpin guidance on this important industry and public health issue. Kinetic models were developed alongside detailed experiments (in which the ester formation reactions were followed with frequent sampling) to inform and direct experimentation before reaching strongly supported conclusions and guidelines for the industry.
The paper is a fine example of a science and risk-based approach, in which "kinetic models are presented that should be of value to process development scientists in designing appropriate controls in situations where risk for sulfonate ester formation does exist".
More at: http://pubs.acs.org/doi/abs/10.1021/op900301n#aff2
Tuesday, March 30, 2010
Monday, March 29, 2010
New solubility prediction utility for solvent selection in DynoChem
Ability to predict solubility of organic solids in common solvents has high value in early phase process development, supporting quick and rational solvent / mixture selection from the myriad of possibilities, consistent with Quality by Design.
This has been an area of intensive research and development in recent years (e.g. for reviews, see references 1, 2 and 3 below). Now, Scale-up Systems has independently developed a new, user friendly solubility prediction method and calculation utility, denoted 'RU' (for Regressed UNIFAC) below; when applied to predict solubility for a variety of common solutes, the method gives results to date of equivalent and sometimes better accuracy when compared to published results using other techniques.
RU builds on the well established UNIFAC and UNIQUAC group contribution approaches, similar to the 'local UNIFAC' method described in reference 1, and requires a small number of experiments to regress the interaction parameters between the solute (e.g. new chemical entity) and the functional groups of solvents in the liquid phase. These results are then leveraged in the utility to select a shortlist of solvent and / or antisolvent candidates.
For reference and illustration purposes, predictions using RU are compared with those of NRTL-SAC below for the solute Cimetidine. The NRTL-SAC predictions are those published in reference 3.
Figure 1: Results of regression to 6 solvents. Both RU and NRTL-SAC have apparent outliers. Because of the solvents (functional groups) excluded from the regression, RU is not applicable to two of the solvents.
Figure 2: Results of regression of RU to 12 solvents compared to NRTL-SAC regression to 6 solvents. RU no longer has significant outliers as all relevant solvent functional groups are covered by the regression.
Figure 3: Results of regression of both RU and NRTL-SAC to 6 solvents, applied to predictions for ethanol/water mixture. Both RU and NRTL-SAC capture the overall shape, with a maximum solubility around 70% ethanol. NRTL-SAC under-estimates and RU over-estimates the maximum solubility.
The new DynoChem utility containing RU will be published shortly and full details will be available to members of DynoChem Resources. (Our new utility for detailed design of crystallizations in late-phase development has been available since December.)
References
1. Peter A. Crafts, The role of crystallization and solubility modeling in the design of active pharmaceutical ingredients, Chapter 2 in Chemical product design : toward a perspective through case studies, By Ka M. Ng, Rafiqul Gani, Kim Dam-Johansen (editors), preview at: http://bit.ly/bPQ4D1
This has been an area of intensive research and development in recent years (e.g. for reviews, see references 1, 2 and 3 below). Now, Scale-up Systems has independently developed a new, user friendly solubility prediction method and calculation utility, denoted 'RU' (for Regressed UNIFAC) below; when applied to predict solubility for a variety of common solutes, the method gives results to date of equivalent and sometimes better accuracy when compared to published results using other techniques.
RU builds on the well established UNIFAC and UNIQUAC group contribution approaches, similar to the 'local UNIFAC' method described in reference 1, and requires a small number of experiments to regress the interaction parameters between the solute (e.g. new chemical entity) and the functional groups of solvents in the liquid phase. These results are then leveraged in the utility to select a shortlist of solvent and / or antisolvent candidates.
For reference and illustration purposes, predictions using RU are compared with those of NRTL-SAC below for the solute Cimetidine. The NRTL-SAC predictions are those published in reference 3.
Figure 1: Results of regression to 6 solvents. Both RU and NRTL-SAC have apparent outliers. Because of the solvents (functional groups) excluded from the regression, RU is not applicable to two of the solvents.
Figure 2: Results of regression of RU to 12 solvents compared to NRTL-SAC regression to 6 solvents. RU no longer has significant outliers as all relevant solvent functional groups are covered by the regression.
Figure 3: Results of regression of both RU and NRTL-SAC to 6 solvents, applied to predictions for ethanol/water mixture. Both RU and NRTL-SAC capture the overall shape, with a maximum solubility around 70% ethanol. NRTL-SAC under-estimates and RU over-estimates the maximum solubility.
The new DynoChem utility containing RU will be published shortly and full details will be available to members of DynoChem Resources. (Our new utility for detailed design of crystallizations in late-phase development has been available since December.)
References
1. Peter A. Crafts, The role of crystallization and solubility modeling in the design of active pharmaceutical ingredients, Chapter 2 in Chemical product design : toward a perspective through case studies, By Ka M. Ng, Rafiqul Gani, Kim Dam-Johansen (editors), preview at: http://bit.ly/bPQ4D1
2. Peter A. Crafts, Pharmaceutical PSE An Industrial Perspective, PSE 2009, Salvador, Brazil: http://www.bit.ly/92nsER
3. Chau-Chyun Chen & Peter A. Crafts, Correlation and Prediction of Drug Molecule Solubility in Mixed Solvent Systems with the Nonrandom Two-Liquid Segment Activity Coefficient (NRTL-SAC) Model, Ind. Eng. Chem. Res. 2006, 45, 4816-4824
Labels:
CQA,
QbD,
Solid-liquid separation,
Solubility
Thursday, March 4, 2010
Generating QbD insight for filtration and centrifugation
Many chemical engineers and chemists will know that filter cakes are often compressible to a high degree and the consequence is that pushing harder (e.g. via higher filtration pressure or faster centrifuge spin speed) leads to remarkably longer cycle times, rather than shorter.
You can explore these effects for a given slurry/cake using templates available in DynoChem Resources. A sample of results to whet your appetite is shown below. Click on each graph for a closer look.
Filtration of a material with compressibility index=2 shows that above a certain pressure, further increases in filtration pressure are counterproductive.
Centrifugation of the same material shows the same trend: beyond a certain spin speed, higher spin speed is counterproductive. Addition time (or rate) is also a factor and slower addition can in fact limit the extension in cycle time seen at higher spin speed.
The above results can be generated easily using the DynoChem menu in Excel to set up and run scenarios covering the factor ranges of interest. Parameter fitting to lab scale filtration data uses the DynoChem Fitting window to characterize the material.
You can explore these effects for a given slurry/cake using templates available in DynoChem Resources. A sample of results to whet your appetite is shown below. Click on each graph for a closer look.
Filtration of a material with compressibility index=2 shows that above a certain pressure, further increases in filtration pressure are counterproductive.
Centrifugation of the same material shows the same trend: beyond a certain spin speed, higher spin speed is counterproductive. Addition time (or rate) is also a factor and slower addition can in fact limit the extension in cycle time seen at higher spin speed.
The above results can be generated easily using the DynoChem menu in Excel to set up and run scenarios covering the factor ranges of interest. Parameter fitting to lab scale filtration data uses the DynoChem Fitting window to characterize the material.
Labels:
Centrifugation,
Design Space,
DOE,
Filtration,
QbD,
Solid-liquid separation