Conclusions

In th is c h a p te r th e fractionation diagram has been shown to be an effective aid in th e selection of an optimum p ro d u ct fraction in chrom atography - e ith e r for an optimum high yield or high p u rity . The su ccessfu l determ ination of th e p ro d u ct fraction is dependent upon deconvolution of th e chromatogram u n d e r examination producing an a cc u ra te model of individual component elution profiles. This is d ep en d en t on th e relativ e peak overlap and peak h eig h ts (see section 2.2.5 and 2.2.6). Additionally, fo r real system s, th e p resence of ex tra peaks (due to contamination or high resolution of minor resolution th an in o th er sep aratio n s) in th e experim ental chromatogram may ad v ersely affect th e accu racy of th e fractio n selection.

A ccurate selection of an optimum high yield fractio n s is d ependent on th e a c c u ra te determ ination of th e s ta r t and end of th e p ro d u ct peak (by deconvolution) w hereas th e accu rate selection of an optimum high p u rity fractio n re q u ire s an a cc u ra te d escrip tio n of all peaks in th e chromatogram so th a t rela tiv e amount of p ro d u ct and contam inant in th e fractio n may be calculated. The la tte r situation is more complex and th u s re q u ire s a h ig h er level of accuracy from th e model chromatogram obtained by deconvolution.

The volume of th e optimum fractio n was found not to be a param eter which could be optimised (ie. for a given optimum p u rity or yield only one volume was available) when selecting th e b e st p ro d u ct fraction. This is tr u e of chrom atographic sep aratio n s w here th e p ro d u ct m aterial is elu ted in one peak. If it is n ecessary to change th is param eter th en it is re q u ire d to re-optim ise th e sep aratio n conditions.

This c h a p te r has dem onstrated th e u tility of combining th e tech n iq u es and algorithm s developed within th is th e s is for th e a t-lin e control of chrom atographic sep aratio n s. C riteria n ecessary for su ccessfu l fractio n selection have been identified and te ste d with both sy n th esised and real chrom atographic data. The control capability offered by th e combination of deconvolution, fuzzy logic peak identification, and fractionation diagram s is sig n ifican tly in advance of existing methods which fail to monitor o r control th e quality of th e p ro d u ct obtained from separations.

start Yes No Yes fraction end > peak end stop No No productivity > threshold Yes fraction start >= product peak end fraction start = product peaks start

store fraction parameters increment fraction end fraction end = fraction start increment fraction start

calculate total & product peak areas for

selected fraction select fraction with highest purity

Figure 4.1 Flow Diagram for the selection of the optimum fraction based on a productivity threshold.

E

■S

è

I

F

A _{Total Protein ( Mass )}

B

Figure 4.2 The Fractionation Diagram

X

Figure 4.3 The Fractionation - Concentration Diagram.

The line from A to B re p re s e n ts an increase in volume only (ie. before th e elution of any material sta rts ). Between B and C, and D and E contaminant protein is eluted (an increase in all th re e param eters can be observed). The product protein is eluted in the central portion of th e line (C to D, - th e increase in total protein is mainly due to th e increase in product protein).

I- I

Total Protein ( Mass )

Figure 4.4 A tie line which represents three fractions with equal purities and purification factors. The fractions are represented by the lines AB, BC and AC. Although they have equal purities the actual composition of the impurities is different.

Ymax Ymin Pmax Vmin Pmin Ymax

Figure 4.5 The Fraction Volume

10 8 y} 6 4 2 0 80 100 0 20 40 60

Total Protein ( Arbitrary Units )

objective function =-1.497 objective function =-7.492 actual peak functions

Figure 4.6

F ra c tio n a tio n cu rv es for th e first G a u ssia n te s t ch rom a to g ra m at tw o o b je c tiv e fu n c tio n va lu es an d th e fr a c tio n a tio n cu rve for th e a c tu a l p eak fu n c tio n s.

5 4 •a X ) 3

1

objective function =-1.444 objective function =-7.485 actual peak functions

Figure 4.7

Fractionation curves for the fourth Gaussian test chromatogram at two objective function values and the fractionation curve for the actual peak functions

10-1 8 — 6 - c s ■e < c

i

1

Elution Volume ( Arbitrai Units )

objective function = -1.497 objective function = -7.492 • actual peak functions

Figure 4.8 The cummulative elution o f product versus elution volume for the deconvolution results and the actual chromatogram for the first Gaussian test chromatogram.

Z) c:' -e < o 2 CL. 2 Q_ 4 3 1 0 5 0 10 15 20 25

Elution Volume ( Arbitrai} Units )

■ objective function = -1.444 ■ actual peak functions

objective function = -7.485

Figure 4.9 The cummulative elution o f product versus elution volume for the deconvolution results and the actual chromatogram for the fourth Gaussian test chromatogram.

4 g. 2 1 0 25 0 3 10 15 20 Total Protein ( mg )

... first egg white chromatogram second egg white chromatogram

fifth egg white chromatogram

Figure 4.10 The fractionation diagrams for the first, second and fifth egg white chromatograms, constructed using deconvolution data

C hapter 5 Discussion

5. Discussion

In th is ch ap ter th e ty p e of chrom atographic sep aratio n which may be adequately analysed (and hence controlled) using tech n iq u es discussed and developed in previous c h a p te rs a re examined - in p a rticu la r th e perform ance of th ese sep aratio n s (ie. th e p u rities and p ro d u ctiv ities which may be achieved).

The effects of linking th e tech n iq u es developed in th is th e sis to g eth er will also be examined, as will th e to tal time req u ire d for th e analyses and how th is would fit into an actual chrom atographic process. Modifications which may be carried out to improve th e tech n iq u es will also be considered.

A comparison with existing work in related areas is also given. 5.1 Separation perform ance

As discussed in th e preceding c h a p te rs th e n a tu re of th e chromatogram is critical in determ ining w hether deconvolution will produce a model chromatogram which adequately d escrib es th e tr u e individual component elution profiles and which may th en be used for p ro d u ct identification and optimum fractio n selection. The param eters which have been found to be im portant in determ ining w hether su ccessfu l chromatogram deconvolution is possible a re th e relativ e peak h eig h ts and th e peak sep aratio n (see section 2.3.10). These may be ex p ressed as a ratio of peak to valley h eig h t (P:V). The re s u lts from c h a p te r 2 indicate th a t a P:V of a t least four is req u ire d fo r one of th e valleys whilst th e ratio for th e o th er valley should not be less th a n one, fo r production of an ad eq u ate model chromatogram. The size of th is ratio for th e two valleys in th is ty p e of model chromatogram give a m easure of th e peak overlap for th e p a rticu la r se t of peak h eig h ts. By an examination of th e relativ e peak overlaps, peak h eig h ts and P:V in th o se te s t chrom atogram s used in section 2.2.6 th e possible quality of sep aratio n (ie. p u r if action factor) th a t is obtainable from th is ty p e of sep aratio n may be determ ined. This examination is c arried out in th e following p a ra g rap h s.

The maximum p ro d u ctiv ities th a t may be obtained from any p a rticu la r chromatogram a re determ ined by th e peak area of th e p ro d u ct peak (for co n stan t peak width). The p u rity th a t may be obtained however, is affected by th e overlap of peaks ie. th e size of th e P:V ratio, in addition to relativ e peak height. As an indication of th e perform ance which is

possible from separation, th e p u rity available a t maximum yield and th e yield a t 100% p u rity , may be used since th ey re p re s e n t extrem es of sep aratio n quality obtainable from a chromatogram.

Table 5.1 Yields for maximum purity and Purities for maximum yield for test chromatograms used section 2.2.6 and Appendix A2

High P u rity / High Yield Yield ( % ) P u rity ( % ) 1st Gaussian te s t chromatogram High p u rity 46.8 > 99.0 High yield 100.0 8.2 2nd Gaussian te s t chromatogram High p u rity 70.0 > 99.0 High yield 100.0 80.0 3rd Gaussian te s t chromatogram High p u rity 54.0 > 99.0 High yield 100.0 49.2 4th Gaussian te s t chromatogram High p u rity 86.0 > 99.0 High yield 100.0 35.6

1st General Exponential Function te s t chromatogram

High p u rity 60.0 > 99.0

High yield 100.0 29.2

2nd General Exponential Function te s t chromatogram

High p u rity 87.0 > 99.0

High yield 100.0 27.2

3rd General Exponential Function te s t chromatogram

High p u rity 45.5 > 99.0

High yield 100.0 11.4

N.B. These fig u res assum e equal extinction coefficients.

from th e te s t chrom atogram s when e ith e r maximum yield (p roductivity) or p u rity a re re q u ire d . The f ir s t Gaussian and th ird General Exponential Function te s t chrom atogram s (see tab le 2.12) a re th e only chrom atograms in table 5.1 which a re not satisfacto rily deconvoluted. Such sep aratio n s would be considered, in general, to be u n sa tisfa c to ry since a low peak sep aratio n is achieved. In su ch cases th e inability of th e deconvolution algorithm to determ ine th e individual component elution profiles from chrom atogram s such as th e f ir s t Gaussian and th ird General Exponential Function te s t chrom atogram s is not a sig n ifican t d isadvantage since th ey re p re s e n t in h ere n tly poor sep aratio n s and would not in general be adopted in a process.

It should be noted th a t th e fig u res in tab le 5.1 a re calculated assum ing equal extinction coefficients for all components. The biological system studied in p revious c h ap ters, hen egg white, contains components with differing extinction coefficients (see tab le 2.10) - one component, Lysozyme, has an extinction coefficient approxim ately twice th a t of o th ers. If it is assum ed th a t th is difference is typical th e effect on th e p u rities th a t may be obtained can be examined for th e high yield fractio n s. The re s u lts a re displayed in Table 5.2.

T ab le 5.2 P u r itie s a t maximum y ie ld s w ith d if f e r in g e x tin c tio n c o e ffic ie n ts Test Chromatogram P u rity at equal E ( table 5.1 ) P u rity with p ro d u ct peak e ^ E * 2 P u rity with p ro d u ct peak E ^ E / 2 1st Gaussian 8.2 4.3 14.1 2nd Gaussian 80.0 66.7 61.5 3rd Gaussian 49.2 32.6 70.0 4th Gaussian 35.6 21.7 52.5 1st General Exponential Function 29.2 17.1 45.2 2nd General Exponential Function 27.2 15.7 42.8 3rd General Exponential Function 11.4 6.0 18.6

The p u rity fig u res for th e te s t chrom atogram s calculated assum ing th e p ro d u ct extinction coefficient is twice th a t of o th er com ponents a re lower th an th o se given in table 5.1. Assuming th a t th e p ro d u ct extinction coefficient is half th a t of o th er components give h ig h er p u ritie s th an those in tab le 5.1. These re s u lts a re due to th e resp ec tiv e in crease and decrease in th e masses of th e p ro d u ct calculated.

Despite th e increase in th e relativ e amounts of contam inants caused by red u cin g th e extinction coefficient of th e p ro d u ct peak n e ith er in th e case of th e f i r s t Gaussian or th e th ird General Exponential Function te s t chrom atogram s does th e p u rity of th e high yield fractio n s exceed tw enty p ercen t. For th e p u rity of th ese frac tio n s to reach th e levels of o th er chrom atogram s (approxim ately fifty p ercen t) th e p ro d u ct extinction coefficients would need to be eig h t times less th an both th e contam inant coefficients. With d iffering extinction coefficients th e peak overlap is not