Characterization of absorption enhancers for orally administered therapeutic peptides in tablet ...

225 Pages·2017·26.38 MB·English

Downloaded from orbit.dtu.dk on: Jan 06, 2023

Characterization of absorption enhancers for orally administered therapeutic peptides
in tablet formulations - Applying statistical learning
Welling, Søren Havelund
Publication date:
2017
Document Version
Publisher's PDF, also known as Version of record
Link back to DTU Orbit
Citation (APA):
Welling, S. H. (2017). Characterization of absorption enhancers for orally administered therapeutic peptides in
tablet formulations - Applying statistical learning. Technical University of Denmark. DTU Compute PHD-2016 No.
429

General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright
owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

 Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
 You may not further distribute the material or use it for any profit-making activity or commercial gain
 You may freely distribute the URL identifying the publication in the public portal

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
Ph.D.Thesis
DoctorofPhilosophy
Characterizationofabsorptionenhancersfororallyadministeredtherapeutic
peptides in tablet formulations
Applyingstatisticallearning
Soren Havelund Welling
KongensLyngby2016
DTUCompute
DepartmentofAppliedMathematicsandComputerScience
TechnicalUniversityofDenmark
Matematiktorvet
Building303B
2800KongensLyngby,Denmark
Phone[phone]
[email protected]
www.compute.dtu.dk
Summary
Todevelopasuccessfulanoralformulationofinsulinfortreatmentoftype-2diabetes
patients would be a great mile stone in terms of convenience. Besides protecting
insulinfromenzymaticcleavageinthesmallintestine,theformulationmustovercome
the intestinal epithelia barrier. Absorption enhancers are needed to ensure even a
few percent of insulin are taken up. In thesis article 1, various methods to measure
the effect of absorption enhancement and enzyme stability of insulin were applied.
The major class of absorption enhancers is surfactant-like enhancers and is thought
to promote absorption by mildly perturbing the epithelial membranes of the small
intestine. The Caco-2 (Carcinoma Colon) cells can grow an artificial epithelial layer,
andareusedtotestthepotencyofnewabsorptionenhancers. Thisprojectwasaimed
to identify new absorption enhancers, that are both potent and sufficiently soluble.
Quantitativestructuralactivityrelationship(QSAR)modelingisanempiricapproach
to learn relationships between molecular formulas and the biochemical properties
usingstatisticalmodels. Apublicdatasettestingthepotencyofabsorptionenhancers
in Caco-2 was used to build a QSAR model to screen for new potent permeation
enhancers. Thesisarticle2containslikelythefirstQSARmodeltopredictabsorption
enhancement. The model was verified by predicting molecules not tested before in
Caco-2. The Caco-2 model overestimates the clinical effect of lipophilic permeation
enhancers. IntheCaco-2modelallreagentsarepre-dissolved,andthereforetheassay
cannot predict critical solubility issues and bile salt interactions in the final tablet
formulation. A QSAR solubility model was built to foresee and avoid slow tablet
dissolution. Due to enzyme kinetics, slow tablet dissolution will allow most insulin
to be deactivated by intestinal enzymes. The combined predictions of potency and
solubility, will likely provide a more useful in-silico screening of potential permeation
enhancers.
Random forest was used to learn relationships between molecular descriptors and
potencyorsolubility. However, unlikemultiplelinearregression, theexplicitlystated
randomforestmodeliscomplex,andthereforedifficulttointerpretandcommunicate.
Any supervised regression model can be understood as a high dimensional surface
connecting any possible combination of molecular properties with a given prediction.
This high dimensional surface is also difficult to comprehend, but for random forests,
itwasdiscoveredthatamethod,featurecontributions,wasespeciallyusefultodecom-
pose and visualize model structures. The visualization technique was named forest
floor and could replace the otherwise widely use technique partial dependence plots,
especially in terms of discovering interactions in the model structure. Thesis article
ii Summary
3 describes the forest floor method. An R package forestFloor was developed to com-
putefeaturecontributionsandvisualizetheseaccordingtotheideasofthesisarticle3.
Betterinterpretationofrandomforestmodelsisanexcitinginterdisciplinaryfield, as
itallowsinvestigatorsofmanybackgroundstofindfairlycomplicatedrelationshipsin
datasetswithoutinadvancespecifyingwhatparameterstoestimate. Forestfloorwas
used to explain how potency and solubility were predicted by random forest models.
Preface
ThisPh.D.thesiswaspreparedatthedepartmentofAppliedMathematicsandCom-
puter Science at the Technical University of Denmark in fulfillment of the require-
mentsforacquiringaIndustrialPh.D.degreeinAppliedMathematicsandComputer
Science.
A number of figures from third party sources have been copied or reproduced in
thesis accordingly to the guide lines Keep your thesis legal [Joh+15]. For any figure
stated as copied, I am not the copy right holder, and I have included the figure as
less than a substantial part of someone others work to make a specific point. Other
figures are either my creations, reproductions and/or copied from public available
sources or mix of all three and can be copied and modified freely.
The results of Caco-2 and Ussing studies of thesis article 1, have been included
before in my master dissertation. However the, enzymatic stability and calcium
electrodestudiesandthepreparationofthemanuscriptwerenotapartofthemaster
thesis project.
Kongens Lyngby, August 23, 2016
Soren Havelund Welling
iv
Acknowledgements
Thanks to Christian Vind for introduction to the Molecular Operating Environment
andforconsultingmeintheearlystageoftheproject. ThankstoSaraØsterBrebbia
(Dirksen) and Sten B. Christensen for collecting permeability data.
ThankstothemanypackageauthorsofR.Thankstothemanypeoplewhocareto
answer questions in forums, write guides, blogs and tutorials. Thanks to github.com,
travis.cl, r-forge.r-project andCRAN forhosting, buildingandtestingtheRpackage
forestFloor.
This industrial PhD program have been sponsered by Innovation Fund Denmark
and the STAR programme at Novo Nordisk.
AspecialthankstomydearsupervisorsHanneHFRefsgaard,LineKClemmensen,
Per B Brockhoff, Stephen T Buckley and Lars Hovgaard. Eventually I will pass on
your advice to others, believing it were my own thoughts.
vi
Contents
Summary i
Preface iii
Acknowledgements v
Contents vii
1 DiabetesType-2andOralInsulin 1
1.1 Diabetes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.1 Blood sugar regulation in diabetic and healthy patients . . . . 2
1.1.2 Treatment and compliance. . . . . . . . . . . . . . . . . . . . . 4
1.2 Drug Development Challenges of Oral Insulin . . . . . . . . . . . . . . 5
1.2.1 Acid resistant coating . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.2 The peptide itself: Size does matter, so does lasting long and
stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2.3 Permeation enhancers to open the epithelial barrier . . . . . . 8
2 Measurepermeationenhancement 11
2.1 Studies to test peptide absorption . . . . . . . . . . . . . . . . . . . . 11
2.1.1 Caco-2 monolayers . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.1.2 Measuring permeability . . . . . . . . . . . . . . . . . . . . . . 13
2.1.3 Calculating permeability . . . . . . . . . . . . . . . . . . . . . . 14
2.2 Mechanisms of absorption enhancement in oral formulations . . . . . . 17
2.2.1 Preventing and measuring enzymatic degradation . . . . . . . . 17
2.3 Article1: Theroleofcitricacidinoralpeptideandproteinformulations:
Relationship between calcium chelation and proteolysis inhibition . . . 18
3 Introductiontotoolsofsupervisedmachinelearning 27
3.1 Supervised machine learning to predict and to learn . . . . . . . . . . 27
3.1.1 Univariate regression . . . . . . . . . . . . . . . . . . . . . . . . 29
3.1.2 Multiple linear regression and interaction terms . . . . . . . . . 29
3.2 Cross-validation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2.1 Segregation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2.2 Unbiased estimation of prediction error . . . . . . . . . . . . . 32
3.2.3 Independent and identical sampling . . . . . . . . . . . . . . . 32