The Application of PAT to Complex Molecule Synthesis

Transcription

1 The Application of PAT to Complex Molecule Synthesis Peter McDonnell PhD, Senior Technical Director, Chemistry and Biotechnology Development, Haverhill, Suffolk, UK Global Innovation Strategy and External Partnerships Manager, Paris, France Heidelberg, ctober 16 th 2014

2 Disclaimer The views expressed in this presentation are those of the author and do not reflect the thinking of Sanofi or its affiliates. The purpose of this presentation is to sponsor debate. The content does not indicate any strategic direction that Sanofi is implementing, unless otherwise stated. Errors and omissions are the author s alone. 2

3 The Application of PAT to Complex Molecule Synthesis Case study: ligonucleotides ligonucleotides are large small molecules that are excluded from current ICH guidelines About 175 oligonucleotides are currently in clinical trials 1 st generation antisense oligonucleotides 2 nd generation antisense oligonucleotides sirna mirna Aptamers LNAs 3

4 Mipomersen structure C 230 H 305 N P 19 S 19 Na 19 MW = Da 2 19 = isomers at P 4

5 Mipomersen (antisense oligonucleotide) X DMT Detritylation X H N N H B Ac Coupling Detritylation NC P S X R n NC DMT P N R B Ac NC + NC N H N N DMT B Ac 1. Sulfur transfer DMT B Ac API 1. Cleavage 2. Purification 3. Deprotection 4. Freeze drying NC P S X R n N 2. Capping NC P X R n N 5

6 QbD and PAT for oligonucleotides Automated synthesis Solid phase (cf peptides) synthesis Excess reagents used to ensure completion of conversion at each step If PAT can be used to prove delivery of correct amidite (base) can this be used in lieu of final testing (at least for sequence)? What about leaks, etc? prove that what enters the column was delivered Can PAT control process? Mipomersen is prepared according to global optimum for oligonucleotides. No large scale manufacture of an oligonucleotide has been required before Could reduce cost by reducing excess reagents, solvent use, recycle loops, etc 6

7 ligonucleotide PAT It is possible despite Dilute solutions onto column Very dilute samples from column Fluorescence of samples (problem with Raman) Standard platform (GE Unicorn) supporting only 12 input signals (solved using 3 rd party integrator) Just monitoring inputs and outputs is insufficient PAT is not just about the use of sensors! QbD requires a much deeper understanding of processes What happens inside synthesis column? Why are excess reagents required? Why does scale up beyond a certain level fail? Why is solid phase peptide synthesis above 60 cm column size performed in stirred beds? 7

8 Magnetic Resonance Imaging All images recorded at Cavendish Laboratories, University of Cambridge and used with permission of Dr Mick Mantle Mipomersen MRIs.mov amidite addition MRI movie-pcwh-jan 2011.avi 8

9 Magnetic resonance imaging 9

10 Challenges Linear (one step at a time) synthesis 99% yield at each step gives: 0.99^80 < 50% overall yield Yield depends critically on good distribution of reagents Ideally we would like plug flow the reality is somewhat different! Swelling & shrinking of bed during synthesis results in poor bed packing 10 G E

11 Electrical Resistance Tomography Column Design P1 P2 P3 P4 11 G E

12 Column Design New column uses the architecture of the existing column as far as practical 12 G E

13 Time :- 1:54:29 13 G E

14 Time :- 1:55:29 14 G E

15 16:56:25 15 G E

16 Zone arrangement Zoned Averages 16 G E

17 Challenges Linear (one step at a time) synthesis 99% yield at each step gives: 0.99^80 < 50% overall yield Yield depends critically on good distribution of reagents Ideally we would like plug flow the reality is somewhat different! Swelling & shrinking of bed during synthesis results in poor bed packing 17 G E

18 Column Design P1 P2 P3 P4 18 G E

19 Column Design New column uses the architecture of the existing column as far as practical 19 G E

20 Time :- 1:54:29 20 G E

21 Time :- 1:55:29 21 G E

22 16:56:25 22 G E

23 Zone arrangement Zoned Averages 23 G E

24 Superimposed spectra of raw Raman data from in line flow cell 24

25 Superimposed spectra of raw data after removal of fluorescence 25

26 27_JUN_Spectra_baseline_corrected_2400_2800.wmv 26

27 Resolution of inlet Raman data 8 components 27

28 Multivariate Curve Resolution of Raman spectra able to see 5 components going on to column 2000 ME C G A C T Spectrum number 28

29 % Conversion of oligonucleotide 1.0 t[2] t[1] SIMCA-P :31:59 (UTC+0) Comp[1] Comp No :31:13 (UTC+0) R2Y(cum) Q2(cum) Comp[2] w*c[1] FlowRate ColumnHeight ColumnDiameter VoidFraction AmiditeConc PecletNumber AmiditeDeliveryTime SolventHeadSpace EndfRecycle Conversion Var ID (Primary) SIMCA-P :33:04 (UTC+0) 29

30 In-line miniature MS Microsaic 3500 MD electrospray MS 30