Introduction to chemical product design design, molecular design problem definition & role of property t modelling d lli Rafiqul Gani CAPEC D Department t t off Chemical Ch i l E Engineering i i Technical University of Denmark DK-2800 Lyngby, y g y Denmark Overview - 1 P bl Problem Definition D fi iti Process Design Raw material Product Design Products Process Fl Flowsheet? h t? Operating condition? Product properties Equipment parameters? Product molecule/ mixture? Product functions? Product Application Design Product Application process? Application conditions? Atoms or molecules Molecule or mixture synthesis? Performance Equipment parameters? Integrate Process-Product & Application Workshop on Product Design, NTUA, March 2011 2 Overview - 2 E Examples l off Product-Process P d tP Integration I t ti Process Design Raw material Product Design Process Fl Flowsheet? h t? Operating condition? Products Equipment parameters? Product properties Product molecule/ mixture? Product functions? Atoms or molecules Molecule or mixture synthesis? •Petroleum P t l blends bl d & petroleum t l products d t Routinely solved! •Polymers & blends Can be solved! •Specialty chemicals (including drugs, ….) Process role? Workshop on Product Design, NTUA, March 2011 3 Overview - 3 E Examples l off Product-Process P d tP Integration I t ti For products F d t such h as drugs, pesticides, food, … delivery and application is important & needs to be validated Product Design Product properties Product molecule/ mixture? Product functions? Product Application Design Product Application process? Application conditions? Atoms or molecules Molecule or mixture synthesis? Performance Equipment parameters? Production process needs to be reliable – first time right is important Workshop on Product Design, NTUA, March 2011 4 Course Content • Different types of design problems •Modelling issues pp •Solution Approaches • Concepts • Single molecule design • Mixture (blend) design Workshop on Product Design, NTUA, March 2011 5 Different Types of Design Problems 1 Design of Molecule and/or Mixture 1. Given, a set of target properties , find molecules or mixtures that match the target properties CAMD & CAMbD solvents, l t fluids, fl id … d drugs, pesticides, ti id aroma, …. Low L Large Value High Production rate Small Easy Finding candidates Difficult Workshop on Product Design, NTUA, March 2011 6 Different Types of Design Problems 1 Design of Molecule and/or Mixture 1. CAMD & CAMbD: Example From a list of 250 solvents, find those that can be used as an oilpaint additive characterized by 25%-, 50%- & 75%-evaporation rate, solubility, viscosity and surface tension. Form the feasible set, find the optimal cost solvent mixture. • Sy Synthesis es s method e od for o mixtures u es • Property models for solvents • Evaporation models for solvents • Cost = f(Ci, xi) Note: Mixture design similar to oil-blend (petroleum, (petroleum edible oil, ….), polymer blend, formulations, …. Workshop on Product Design, NTUA, March 2011 7 Different Types of Design Problems 1 Design of Molecule and/or Mixture 1. Given, a set of target properties , find molecules or mixtures that match the target properties CAMD & CAMbD solvents, l t fluids, fl id … d drugs, pesticides, ti id aroma, …. Issues & Needs • Generate feasible candidates (synthesis method & tool) • Estimate E ti t target t t properties ti (predictive ( di ti property t models) d l) • Evaluate performance (application process models) A h i ett al, Achenie l Computer C t Aided Aid d M Molecular l l D Design: i Th Theory & Practice, P ti CACE CACE-12, 12 2003 Workshop on Product Design, NTUA, March 2011 8 Different Types of Design Problems 2 Product 2. Product-Process Process Evaluation Given, a list of feasible candidates (product and/or process), the objective bj ti is i to t identify/select id tif / l t the th mostt appropriate i t product-process d t based on a set of performance criteria. For product design, design this problem is similar to CAMD but without the generation step; also, similar to CAMbD where additives are added to a product to significantly enhance the performance of the product Examples • Select the optimal combination of pesticide and surfactants that match the needs of specific plants • Select the optimal combination of drug/pesticide, solvent and polymeric microcapsule for controlled release of the product • Increase the yield of a product by hybrid process operation Workshop on Product Design, NTUA, March 2011 9 Different Types of Design Problems 2 Product 2. Product-Process Process Evaluation Simple Example: From the derivatives of Barbituric acid, identify the candidate did t that th t is i required i d in i the th smallest ll t amountt (C) tto produce d a 1:1 complex with protein (binding to bovine serum albumin) – calculated as a function of octanol-water partition coefficient of the candidate compounds: Log10(1/C) = 0.58 log10P + 0.239 • Synthesis tool for isomer generation • Property model for Log10P • Verify solubility in water Note: How is the backbone (barbituric acid) identified and how is the relation (drug activity) between C vs log10P found? Workshop on Product Design, NTUA, March 2011 10 Different Types of Design Problems 2 Product 2. Product-Process Process Evaluation Process Analytical Technology – PAT To ensure final product p quality! Estimate critical iti l scientific & regulatory parameters early! A. S. Hussain, CDER, FDA, 2002 Workshop on Product Design, NTUA, March 2011 11 Different Types of Design Problems 3 Design of Product 3. Product-centric centric Process Problem Characteristics • • • • • Products are consumer products such as soap, detergents, fragrances, food, skin-care ki … Complex chemical systems involved Life of the product is getting shorter Volume/cost relationship p needs to be improved by selling more Low-cost chemical routes are needed (better catalysts, faster reactions, increased yields, …) Issues & Needs Algorithm for generation of process flowsheet fl h t or batch recipe Process-operation models for verification by simulation Property models for product/process analysis Workshop on Product Design, NTUA, March 2011 12 Framework for Integrated Approach - 1 Integration of Product-Product Product Product Design hwax hcuticle Workshop on Product Design, NTUA, March 2011 13 Role of property modelling Prediction P di ti and d measurementt off properties for product design Workshop on Product Design, NTUA, March 2011 14 Motivation – I: Chemicals based products Example-1: Design of a consumer product an insect repellent lotion What does it involve? Determine a liquid formualtion that is effective against mosquito, it h has a pleasant l t smell, ll a good d skin ki feeling and is a water-based product (active ingredient water ingredient, water, solvents solvents, additives) Properties?The above product attributes, when translated, translated means that target values on the following properties need to be p time,, phase p split, p , matched: 90% evaporation solubility, viscosity, molar volume, toxicity, etc., & cost Workshop on Product Design, NTUA, March 2011 15 Motivation – II: Chemical processes E Example-2: l 2 Design D i off a chemical h i l process What does it involve? Combine different unit operations to convert raw materials (chemicals) to desired products (chemicals) in a sustainable, efficient and cost effective manner manner. Properties? Wide range of properties needed to establish t bli h the th chemical h i l product d t (pure ( compound d and/or d/ mixture properties) and the process (how the chemical properties change under conditions of temperature, pressure and/or composition). Monomer A Separator effluent Monomer B Solvent (S) Initiator (I) MIXER Transfer agent (T) Inhibitor (Z) REACTOR SEPARATOR Copolymer product Workshop on Product Design, NTUA, March 2011 16 Challenges & Issues Wide range of products & processes must be handled Nutrition ..... Fertilizers Vitamins Healthcare Electronics Chemical Processes Communications, Entertainment Fuel Mobility Pharmaceuticals Plastics Colorants and Detergents coatings Clothing Workshop on Product Design, NTUA, March 2011 Housing 17 Challenges & Issues How tto provide H id the th necessary property t values l in i a fast, efficient and reliable way? Software tools: models; algorithms; databases Chemical systems: organic; polymers; electrolytes; .. Properties: pure compound & mixture i t Workshop on Product Design, NTUA, March 2011 18 Challenges & Issues Properties for product-process design* Databases: collection of measured data (there are gaps in the database) Property models: fillin the gaps (available models are not perfect ) *Synthesis: generation-screening of alternatives (fast, qualitatively correct predictive models are needed) *Design/analysis: selection-validation (reliable, quantitatively correct models are needed) Workshop on Product Design, NTUA, March 2011 19 Role of property models Property models Monomer A Separator effluent Monomer B Log Pi = Ai + [Bi/(Ci + T)] Property models are embedded into process models Process models Solvent (S) Initiator (I) MIXER dm dt T Transfer f agent (T) Inhibitor (Z) REACTOR i f in ,i f o u t ,i r ( m , T , P )V ; i 1 , N C SEPARATOR Operation models Process models Property-kinetics models Cost models M Models s for su ustaina ability metrics Models fo or ntal envirronmen im mpact Copolymer product Workshop on Product Design, NTUA, March 2011 Formulation process model Product evaluation model 20 Property Modelling Concept Predictive: Group contribution plus atom connectivity y for p primary yp pure component p properties and mixture properties (GE-models*, viscosity, y surface tension, EOS ((PC-SAFT)) ...)) Correlated: Secondary pure component properties (Antoine (Antoine, DIPPR DIPPR-functions, functions equations of state, ....) and mixture properties (GE-models**, viscosity surface tension viscosity, tension, EOS (SRK (SRK, PR) PR), ...)) * UNIFAC ((different versions)) ** NRTL, UNIQUAC, Wilson, .... Workshop on Product Design, NTUA, March 2011 21 Property Model Development – GC concept Organic Chemical & Polymer Properties 1st-order: CH3 , CH2, CH, OH, COOH, CH3CO, …. 3 5 1 2nd-order: (CH3)2CH, CH .. 3rd-order: HOOC-(CHn)mCOOH (m>2, n in 0..2) groups g p 4 2 v 0 v 1 v 2 connectivity y index REPRESENTING DATA SET AS GROUPS & ATOMS EXPERIMENTAL DATA SET COLLECTION Workshop on Product Design, NTUA, March 2011 22 Property Model Development Organic Chemical & Polymer Properties DETERMINING ATOM or GROUP CONTRIBUTIONS MARRERO O / GANI G GC-method GC et od (2001) ( 00 ) f(y) = ∑niCi + w∑mjDj + z∑okEk + F MINIMIZATION OF SUM OF SQUARED RESIDUALS ( experimental – estimated) Atom-CI to C Model ode (2006) ( 006) f(y)=∑aiAi + b(vχ0) + 2c(vχ1) + 2c(vχ2) + D Note: f(y) is function of property x; example f(y) = Exp(Tb /tbo) MODEL DEVELOPMENT REPRESENTING DATA SET AS GROUPS & ATOMS EXPERIMENTAL DATA SET COLLECTION Workshop on Product Design, NTUA, March 2011 23 Property prediction (pure component) Workshop on Product Design, NTUA, March 2011 24 Property prediction (pure component) Workshop on Product Design, NTUA, March 2011 25 Property prediction (pure component) Equations of state: SRK, PR, PCSAFT, .. SAFT Workshop on Product Design, NTUA, March 2011 26 Property prediction: organic chemicals & polymers Modelling options for primary properties GC model + connectivity indices Gcplus concept f (Y ) NiCi f (Y ) higher order * i Generation G ti off missing groups for pure compound GC models[1] GC+ models GC model + experimental data Group contribution based models GC model (MG, CG, JR, W, …) Gani, R.; Harper, P.M.; Hostrup, M. Automatic Creation of Missing Groups through Connectivity Index for Pure-Component Property Prediction. Prediction Ind Ind. Eng Eng. Chem. Chem Res., Res 2005, 2005 44, 44 7269 7269. GC model + force field (MM2) Higher order models for structure distinction (cistrans isomers) Workshop on Product Design, NTUA, March 2011 27 GC+ concept for mixture property prediction UNIFAC GC models Compound 1 Compound 2 lni ln COM i lni lni RES RES2 1. Given: Compounds, compositions & temp. 2. Represent molecules with 1st-order 1st order & 2nd2nd order groups 3. Retrieve group parameters 4. Calculate activity coefficients in liquid phase 5. ....... Equlibrium q systems: y VLE, LLE, SLE, VLLE, SLLE, SLVE Workshop on Product Design, NTUA, March 2011 28 The UNIFAC-CI Concept Methodology to fill the blanks in the UNIFAC matrix A lot of gaps in th UNIFAC matrix! the ti ! VLE, SLE, LLE experiments to collect data Filling the UNIFAC matrix jjust with the published VLE, SLE and LLE data GC model for Atom and bond Based information Mixtures (UNIFAC) (connectivity indices) Sometimes time consuming and expensive and even not possible By using just atom and bond information Of the groups describing g the molecules •Without any extra experimental data •With reasonable accuracyy UNIFAC-CI Gonzales et al , Prediction of Missing UNIFAC Group-interaction, Parameters through connectivity indices, AIChE J (2006); FPE (2010-submitted) Workshop on Product Design, NTUA, March 2011 29 Predict Missing UNIFAC Group Interaction Parameters CH2 C=C C≡C ACH ACCH2 OH CH30H ACOH DOH H2O CH2CO CHO Furfural CCOO HCOO CH2O COOH CNH2 CNH (C)N3 CCN ACNH2 Pyridine CNO2 ACNO2 DMF ACRY CF2 ACF CCl CCl2 CCl3 CCl4 ClCC ACCl Br I CS2 CH3SH DMSO COO SiH2 SiO NMP CClF CON OCCOH CH2S Morpholine Thiophene C,O,H atoms (current state) adding N atoms adding F atoms adding Cl atoms adding Br atoms adding I atoms adding S atoms adding Si atoms Potentially all the group g interactions parameters could be covered using UNIFAC-CI ! adding Cl & F atoms Workshop on Product Design, NTUA, March 2011 30 Property prediction: UNIFAC-CI (VLE) MethylcyclopentaneBenzene at 298 K Ethanol-Hexane at 473 K Workshop on Product Design, NTUA, March 2011 31 Property prediction (mixture): UNIFAC-CI Example: Generation of missing parameters for solubility predictions of pharmaceutical active ingredients Workshop on Product Design, NTUA, March 2011 32 Property prediction software S it d ffor product-process Suited d t d i / design/analysis l i MoT Introduce new models to the system Workshop on Product Design, NTUA, March 2011 33 Database development *K Knowledge l d representation t ti (developed (d l d ontology) t l ) •Classes ((chemicals;; AIs;; Aromas;; Solvents;; ....)) •Sub-classes (organic, polymers, inorganic, ..) •Instances I t (alcohols, ( l h l ketones, k t paraffins, ffi ..)) •Objects-A j (property (p p y types: yp primary, p y, ...)) •Frames (property values: Tb, Tc, Tm, ...) •Objects-B (applications: solvents, ....) * Search engine g for data-retrieval ((forward and/or reverse) R. Singh et al. ”Design of an efficient ontological knowledge based system for selection of process monitoring and analysis tools”, ” C Computers & C Chemical Engineering, 2010 Workshop on Product Design, NTUA, March 2011 34 Database: Contents & features • Pure compound data: primary properties • Binaryy mixture data: VLE and infinite dilution ((in water) • Solid solubility data (C4 – C60) • Special databases • Solvents • Active ingredients • Aroma o a • Agents (neutralizers, emulsifiers, ....) • Search engines & data retrieval interface Example of a lipid database and associated property models (Diaz-Tovar (Diaz Tovar et al al., FPE FPE, 2010) Workshop on Product Design, NTUA, March 2011 35 Hierarchy of property models U a more predictive Use di ti model d l tto predict di t th the missing i i parameter t Correlations Pi = Ai + [Bi/(Ci + T)] Molecular CH3-; -CH2-; -OH; ….. Zc = (Pc*Vc)/(83.14*Tc) Groups C, H, O, N, S, …. Tb = 222.543*log(Sum.Groups.I + Sum.Groups.II + Sum.Groups.III) Atoms Use the smaller scale (atoms) model to predict the missing parameters of the larger scale (GC) P=∑niPi + b(vχ0) + 2c(vχ1) Micro Accuracy (verification) Predictive power (design) Workshop on Product Design, NTUA, March 2011 36 Increasing application range of property models Accuracy y (q (quantitative)) Pi = Ai + Bi + Ci + Di Property Models E Electrolyte (simple small molecules & mixtures) Higher-order models (plus Poly ymers gg more complex molecules & mixtures)) bigger GC+ models (plus isomers & more complex GC molecules & systems) Orrganic M lti Multi-scale l models d l (plus more complex systems and structured chemicals) A li ti range (qualitative) Application ( lit ti ) Workshop on Product Design, NTUA, March 2011 37 Molecular Design (single molecule) Workshop on Product Design, NTUA, March 2011 38 Chemical product design framework Essential, E ti l desirable and EH&S properties Build molecules & mixtures; verify their properties Product & process evaluation Identify processes that can manufacture the product sustainably 1. Define needs; 2. Design products; 3. Design processes; 4. Evaluate process-product Workshop on Product Design, NTUA, March 2011 39 General Product Design Problem as MINLP MINLP means Mixed Integer Non Linear Programming – optimization problems formulated as MINLP contains integer g ((Y)) and continuous variables ((x, y) as optimization p variables, and, at least the objective function and/or one of the constraints is non-linear. Fobj = min {CTY + f(x, y, u, d, θ ) + Se + Si + Ss + Hc + Hp} ( , x,, y, d,, u,, θ)) P = P(f, Process*/product p model 0 = h1(x, y) Process/product constraints 0 g1(x, ( u, d) “Other” (selection) constraints 0 g2(x, y) B x + CTY D Alternatives (molecules; unit operations; mixtures; fl flowsheets; h t ….)) Workshop on Product Design, NTUA, March 2011 40 2. Product Design: CAMD Framework • Solutions of CAMD problems are iterative iterative. Candidate • Problem selection formulation controls the success. • Essential E ti l qualities vs. Finish ((un)desirable ) qualities. • Connectivity with external t l tools, t l data and methods. (final verification/ comparison p using rigorous simulation and experimental procedures) Start No suitable solutions (due to performance, economic or safety concerns) Promising candidates have been identified Problem formulation (identify the goals of the design operation) Method and constraint selection Result analysis and verification (specify the design criteria based on the problem formulation) (analyse the suggested compounds using external tools) Workshop on Product Design, NTUA, March 2011 CAMD Solution (identify compound having the desired properties) 41 CAMD General Problem Solution 3 3-stage Iterative i Solution S i Approach A 1. Set Goals ((Pre-Design g Stage) g ) – Define Needs • • • Essential properties Desirable properties Safety Environmental Safety, Environmental, etc. etc 2. Design/Selection (Design Stage) – Generate Alternatives • • • Search S h through th h database d t b Iteratively build molecules/mixtures, comparing to goals • Generate set of feasible molecules/mixtures Simultaneously build and evaluate molecules/mixtures • Identify optimal molecule/mixture • Identify y feasible set 3. Analyze, verify and select (Post-Design Stage) - Final Selection (process/operation constraints) Workshop on Product Design, NTUA, March 2011 42 CAMD Problem Solution: Database Search Problem: Find solvents that have 19.5 > Sol Par < 20.5 Solution: S l ti U a search Use h engine i within ithi a database d t b tto identify the set of feasible molecules Workshop on Product Design, NTUA, March 2011 43 CAMD Problem Solution: Enumeration Multi-level generate & test according to a predefined sequence Pre-design g "I want acyclic alcohols, ketones, aldeh des and ethers aldehydes with solvent properties similar to Benzene" Interpretation to input/constraints Design g ((Start)) A set of building blocks: CH3, CH2, CH, C, OH, CH3CO, CH2CO, CHO, CH3O CH2O CH3O, CH2O, CH CH-O O + A set of numerical constraints A collection of group vectors like: 3 CH3, 1 CH2, 1 CH, 1 CH2O All group vectors satisfy constraints Design (Higher levels) 2.order group CH2 CH3 CH2 O CH3 CH CH3 CH3 Group from other GCA method CH2 CH3 CH3 CH O CH2 Refined property estimation. Ability to estimate additional properties or use alternative methods. Rescreening against constraints. Workshop on Product Design, NTUA, March 2011 Start of Post-design CH2 CH3 CH2 O CH3 CH CH3 CH3 CH2 CH3 CH3 CH O CH2 44 Example of problem solution with ProCAMD Workshop on Product Design, NTUA, March 2011 45 Integrated Product-Process Design Problem: A process stream of 50 mole% Acetone and 50 mole% Chloroform at 300K, is to be separated. Separation p techniques q considered: h Adsorption (liquid, gas) Crystallization Desublimation Distillation – simple Distillation – extractive Distillation with decanter Liquid-liquid extraction Flash/evaporation Membrane (g (gas,, liquid) q ) Microfiltration Partial condensation N external No t l medium di known k Binary ratios of properties identify the following alternatives Separation techniques: Distillation – simple Distillation – extractive Distillation – azeotropic Liquid extraction Pressure swing Note: Acetone-chloroform forms a high boiling azeotrope that is ppressure sensitive Workshop on Product Design, NTUA, March 2011 46 Pressure dependence Workshop on Product Design, NTUA, March 2011 47 Solvent design sub-problem • • • • • CAMD problem: 340 < Tb < 420 Selectivity y > 3.5 Solvent power> 2.0 No azeotropes Solution: 1-Hexanal y p y ether Methyl-n-pentyl (Benzene) – Number of compounds designed: 47792 Number of compounds selected: 53 – Number of isomers designed: 528 Number of isomer selected: 23 – Total time used to design: 57.01 s Workshop on Product Design, NTUA, March 2011 48 Phase behaviour Workshop on Product Design, NTUA, March 2011 49 Problem formulation & Solution Objective function: Maximize Profit = Earnings – Solvent cost – Energy costs Constraints: Results: Acetone purity > 0.99 S l Solvent t S l Solvent t Chloroform purity 0.98 flow>rate 1-hexanal 0.082 kmol/hr Reflux R fl Reb. 1 0.45 Reflux R fl Reb. 2 0.65 Workshop on Product Design, NTUA, March 2011 Objective Obj ti function 2860.51 $/hr 50 Product Recovery/Purification Combine CAMD, ProPred, Solubility tools & Database To solve following problems – Given, the molecular description of a pharmaceutical product, find solvents needed a) in its production; b) in its formulation Only problem a) to be highlighted Solution strategy: gy Define CAMD p problem to ggenerate solvent candidates; verify performance through solubility calculations; check database to verify predictions Workshop on Product Design, NTUA, March 2011 51 Product Recovery: Crystallization Chemical product: Ibuprofen Consider C id cooling li as well ll as drowning-out crystallization Find solvents, anti-solvents & their mixture that satisfy the following: Potential recovery > 80% Solubility parameter > 18 MPA1/2 (or > 30) H d Hydrogen b di solubility bonding l bilit parameter t > 9 MPA1/2 (or ( > 24) Tm < 270 K; Tb > 400 K; -log (LC50) < 3.5 Solution strategy: Define CAMD problem to generate solvent candidates; verify performance through solubility calculations; check database to verify predictions Workshop on Product Design, NTUA, March 2011 52 Product Recovery: Crystallization Chemical product: Ibuprofen Consider C id cooling li as well ll as drowning-out crystallization Workshop on Product Design, NTUA, March 2011 53 Product Recovery: Crystallization Chemical product: Ibuprofen Consider cooling as well as drowning-out crystallization Workshop on Product Design, NTUA, March 2011 54 Case Study – Polymer Design: MINLP Applied to CAMD Polymer design problem where polymer repeat units with 2-4 different groups and 2-9 total number of groups in the repeat unit structure are being sought F bj = [(Pi – Pi*)/Pi*]2 Fobj s.t. Tg 373 K Density 1.5 1 5 g/cm3 Water absorption = 0.005 g H2O/g polymer) -CH2- ; -CO- ; -COO- ; -O- ; -CONH- ; -CHOH-; -CHCL- Workshop on Product Design, NTUA, March 2011 55 Case Study – Polymer Design: MINLP Applied to CAMD Fobj = [(Pi – Pi*)/Pi*]2 select property models st s.t. Tg = [ (Yknk)]/ [ (Mknk)] (373 20) K Density = [ (Mknk)]/ [ (Vknk)] (1.5 0.1) g/cm3 W t absorption Water b ti = [ (Hknk)]/ [ (Mknk)] (0.005 0.0005)) g H2O/gg p polymer) y ) -CH2- ; -CO- ; -COO- ; -O- ; -CONH- ; -CHOH-; -CHCLNote: objective function and constraints are non linear; nk, the optimization variables are integer (0-9) Workshop on Product Design, NTUA, March 2011 56 Case Study – Polymer Design: MINLP Applied to CAMD Fobj = [(Pi – Pi*)/Pi*]2 st s.t. 353 K < Tg = [ (Yknk)]/ [ (Mknk)] < 393 K 1.4 < = [ (Mknk)]/ [ (Vknk)] < 1.5 g/cm3 0 0045 < W = [ (Hknk)]/ [ (Mknk)] < 0.0055) 0.0045 0 0055) g H2O/g polymer) 2 yj 3 2 yknk 9 ; yj : 0 or 1 for j=1,7 add structural constraints -CH2CH2 ; -COCO ; -COOCOO ; -OO ; -CONHCONH ; -CHOH-; CHOH ; -CHCLCHCL Note: objective function and constraints are non linear; nk, the optimization variables are integer (0-9) (0 9) Workshop on Product Design, NTUA, March 2011 57 Case study – Polymer Design: Reformulate - I Fobj = [(Pi – Pi*)/Pi*]2 st s.t. 353 K < Tg = [ (Yknk)]/ [ (Mknk)] < 393 K 1.4 < = [ (Mknk)]/ [ (Vknk)] < 1.5 g/cm3 0 0045 < W = [ (Hknk)]/ [ (Mknk)] < 0.0055) 0.0045 0 0055) g H2O/g polymer) 50 < (Mknk) < 100 2 yj 3 add a new constraint ; yj : 0 or 1 for j=1,7 2 yknk 9 -CH2- ; -CO- ; -COO- ; -O- ; -CONH- ; -CHOH-; -CHCLWorkshop on Product Design, NTUA, March 2011 58 Case study – Polymer Design: Reformulate – II (MILP) reformulate Fobj & constraints Min Fobj = M st s.t. 18.7 < TgM= [ (Yknk)] < 37.3 K g/mol 50/1.4 < M/ = [ (Vknk)] < 100/1.5 g/cm3 0 0045*50 < W M = [ (Hknk)] < 0.0055*100 0.0045*50 0 0055*100 g H2O/g polymer) LB1 yj UB1 ; yj : 0 or 1 for j=1,7 LB2 yknk UB2 -CH2CH2 ; -COCO ; -COOCOO ; -OO ; -CONHCONH ; -CHOH-; CHOH ; -CHCLCHCL 14 ; 28 ; 44 ; 16 ; 43 ; 30 Workshop on Product Design, NTUA, March 2011 ; 48.5 59 MINLP Applied to CAMD: Exercise 2 G Generate t polymers l that th t match t h the th ffollowing ll i target: t t Total number of different main groups: 5 for Problem 1 4 for Problem 2 4 for Problem 3 2 yjnj 5 ; yj : 0 or 1 for j=1,4 ; yk: 0 or 1 for k=5,8 m=(v ( j -2)n ) j 0 yknk m Workshop on Product Design, NTUA, March 2011 60 MINLP Applied to CAMD: Exercise 3 G Generate t molecules l l that th t match t h the th following f ll i target: t t Min f=Hfus 180 < Tm < 225 K 400 < Tb < 440 K -CH3 ; -CH2NO2; -CH3CO; -OH; -CH2=CH -CH2-CHCH yjnj = 4 ; yj : 0 or 1 for j=1,7 2 vjnj 3 ; yj : 0 or 1 for j=1,5 0 yknk 2 ; yk: 0 or 1 for k k=6 6 0 ylnl 1 ; yl: 0 or 1 for l=7 Workshop on Product Design, NTUA, March 2011 61 MINLP Applied to CAMD: Exercise-4 G Generate t molecules l l that th t match t h the th following f ll i target: t t Min f= Cixi LB1 < Tm = Tmi xi < UB2 LB2 < Vb = f(T, f(T Tc, Pc, Zc, x) < UB2 Comp1, Comp2, Comp3, …………..CompN yj= 2 ; yj : 0 or 1 for jj=1 1,N N 0 < x1 < 1 0 < x2 < 1 x1 + x2 = 1 Workshop on Product Design, NTUA, March 2011 62 Exercise Problems Workshop on Product Design, NTUA, March 2011 63 Tutorial Exercises • Analyze the drug Ibuprofen (define the needs d ffor iimproving i process efficiency ffi i in i extracting the product from solution in the reactor) – use ProPred • Use ProCamd to match the list of solvents and anti-solvents for Ibuprofen • Solve S l the h organic i synthesis h i problems bl with ih ProCamd See slides 53-55 Workshop on Product Design, NTUA, March 2011 64 Exercises • Analyze the drug Ibuprofen (define the needs for improving p gp process efficiency y in extracting the product from solution in the reactor) – use ProPred – Search compound in database • start ICAS • click on M, select basic search, search with CASnumber b (015687 (015687-27-1), 27 1) copy SMILES – Start ProPred • import SMILES, analyze properties Workshop on Product Design, NTUA, March 2011 65 Exercises • Use ProCamd to match the list of solvents and anti-solvents for Ibuprofen – Formulate problems (solvent & anti anti-solvent) solvent) see slides 53-55) • Start ProCamd – Specify problem (ProCamd manual) – Start execution – Analyze results Workshop on Product Design, NTUA, March 2011 66 Other chemical product design problems – 1a Problem Statement-1: From the derivatives of Barbituric acid, identify the candidate that is required in the smallest amount (C) to produce a 1:1 complex with protein (binding to bovine serum albumin) Barbituric Acid (000077-02-1) Workshop on Product Design, NTUA, March 2011 67 Other chemical product design problems – 1b Problem Statement Statement-1: 1: From the derivatives of Barbituric acid, identify the candidate that is required in the smallest amount (C) to produce a 1:1 complex with protein (binding to bovine serum albumin)) Barbituric Acid (000077-02-1) Calculate C as a function of octanol-water partition coefficient of the candidate compounds: Log10(1/C) = 0.58 log10P + 0.239 Workshop on Product Design, NTUA, March 2011 68 Other chemical product design problems – 1c Problem Statement-1: From the derivatives of Barbituric acid, identify the candidate that is required in the smallest amount (C) to produce a 1:1 complex p with protein p (binding ( g to bovine serum albumin)) Barbituric Acid (000077-02-1) (000077 02 1) Calculate C as a function of octanol-water partition coefficient of the candidate compounds: Log10(1/C) = 0.58 0 58 log10P + 0.239 0 239 Question: How is the backbone (barbituric acid) identified and how is the relation ((drugg activity) y) between C vs log g10P ffound? Workshop on Product Design, NTUA, March 2011 69 Other chemical product design problems – 1d Problem Statement-1: From the derivatives of Barbituric acid, identify the candidate that is required in the smallest amount (C) to produce a 1:1 complex p with protein p (binding ( g to bovine serum albumin)) Barbituric Acid (000077-02-1) (000077 02 1) Calculate C as a function of octanol-water partition coefficient of the candidate compounds: Log10(1/C) = 0.58 log10P + 0.239 Solution Steps: Generate candidates; calculate LogP; calculate C; order solutions w.r.t. w r t C to identify the best candidate; check for solubility in water Workshop on Product Design, NTUA, March 2011 70 Other chemical product design problems – 1e Problem Statement-1: From the derivatives of Barbituric acid, identify the candidate that is required in the smallest amount (C) to produce a 1:1 complex p with protein p (binding ( g to bovine serum albumin)) Solution Steps •Generate candidates (check database) •Use ProPred to calc LogP and LogWs •Calculate C Barbituric Acid (000077-02-1) (000077 02 1) •Order solutions with respect p to C Calculate C as a function of octanol-water partition coefficient of the candidate compounds: Log10(1/C) = 0.58 log10P + 0.239 Candidates (CAS Numbers) : 061346-87-0; 000076-94-8; 091430-64-7; 001953-33-9; 007391-69-7; 090197-63-0; 017013-41-1; 027653-63-0 Workshop on Product Design, NTUA, March 2011 71 Other chemical product design problems – 2a Problem Statement-2 (Design of backbones & termination of backbones): In this problem, we will start with ProPred, take a known molecule (for example, Corticosterone), use the features in ProPred to create free attachments in the molecule (that is, create a backbone). The Backbone is then transferred to ProCamd, where terminated structures are generated (see page 53 of tutorial document) Step 1: Start ProPred from ICAS Step 2: Click on (database) , click on “Find CAS”, type the CAS Number for Corticosterone (00005-22-6), select the compound by clicking on OK Step 3: ProPred draws the molecule and predicts the properties Step 4: Remove the OH group connection from the molecular structure at the two locations Step 5: Launch ProCamd from ProPred from the tools menu in ProPred. ProPred Step 6: Fill-out the necessary problem definition pages in ProCamd (general problem control, non-temperature dependent properties) Step 7: Terminate the backbone by clicking on “GO”. Step 8: Save the backbone-termination problem as “save ProPred backbone problem” under the file menu. To use this file later, the saved backbone file must be loaded from the file menu. Workshop on Product Design, NTUA, March 2011 72 Other chemical product design problems: 2b Problem Statement Statement-2 2 (Construct backbones & terminate with ProCamd): Use Corticosterone as an example Step 2 St 2: Cli Click k on (d (database) t b ) , click li k on “Find “Fi d CAS”, CAS” ttype the CAS Number for Corticosterone (00005-22-6), select the compound by clicking on OK Workshop on Product Design, NTUA, March 2011 73 Other chemical product design problems: 2c Problem P bl Statement-2 St t t2 (Construct backbones & terminate with ProCamd): Use Corticosterone as an example Step 3: ProPred draws the molecule and predicts the properties Workshop on Product Design, NTUA, March 2011 74 Other chemical product design problems: 2d Problem P bl Statement-2 St t t2 (Construct backbones & terminate with ProCamd): Use Corticosterone as an example Step p 4: Remove the OH group connection from the molecular structure at the two locations Backbone with two-free attachments. Note that as ProCamd generates g molecular structures only with the Constantinou & Gani groups, it must be possible for the Constantinou & Gani method to represent the backbone. Otherwise, ProPred will not launch ProCamd Workshop on Product Design, NTUA, March 2011 75 Other chemical product design problems: 2e Problem P bl Statement-2 St t t2 (Construct backbones & terminate with ProCamd): Use Corticosterone as an example Step 6: Fill-out the necessary problem definition pages in ProCamd (general problem bl control, t l nontemperature dependent properties) Workshop on Product Design, NTUA, March 2011 76 Other chemical product design problems: 2f Problem P bl Statement-2 St t t2 (Construct backbones & terminate with ProCamd): Use Corticosterone as an example Step 7: Terminate the backbone by clicking on “GO”. Note that for backbone termination step, the options for P P d properties ProPred ti and d database d t b search cannot be used. Workshop on Product Design, NTUA, March 2011 77 Other chemical product design problems: 3a Problem Statement Statement-3 3 (Generate & terminate backbones):In this problem, problem we will first generate a backbone with ProCamd using only the C & H atoms and with 1 free-attachment in the backbone. We will then go to ProPred to draw the molecular o ecu a st structure uctu e a and d work o o on tthe e st structure uctu e without t out terminating te at g the t e structure. We will then launch ProCamd from ProPred to find the final terminated structure through ProCamd (see pages 53-60 in tutorial document) Step 1: Start ProCamd and specify the following backbone generation problem. Step 2: Run ProCamd to generate the backbone alternatives. Note that for the backbone generation, “isomer” generation is not allowed. For the problem formulated in step 1, the following result is obtained. Step 3: Click on ProPred to transfer the backbone structure. structure Propred draws the structure and calculates all properties, as shown below. Step 4: Launch ProCamd from ProPred from the “tools” menu A new ProCamd application is opened. ProPred sends back the same backbone structure that ProCamd generated. Step 5: Complete the backbone termination problem with ProCamd. ProCamd Workshop on Product Design, NTUA, March 2011 78 Other chemical product design problems: 3b Problem Statement3 (Generate backbone with ProCamd & terminate manually with ProPred) Step 1: Start P C d and ProCamd d specify the following backbone generation p problem. Note: at this stage specification of any other property is not necessary Workshop on Product Design, NTUA, March 2011 79 Other chemical product design problems: 3c Problem Statement-3 (Generate backbone with ProCamd & terminate manually with ProPred) Step 2: Run ProCamd to generate the backbone alternatives. l i Note N that h for f the backbone generation, “isomer” generation is not allowed. For the problem formulated in step 1, the following g result is obtained. Workshop on Product Design, NTUA, March 2011 80 Other chemical product design problems: 3d Problem Statement-3 (Generate backbone with ProCamd & terminate manually with ProPred) Step 3: Click on ProPred to transfer the backbone structure. ProPred d draws the th structure and calculates all properties, as shown below. Workshop on Product Design, NTUA, March 2011 81 Other chemical product design problems: 3e Problem Statement-3 (Generate backbone with ProCamd & t terminate i t manually ll with ProPred) Step 4: Launch ProCamd from ProPred from the “tools” menu A new ProCamd application is opened. ProPred sends back the same backbone structure that ProCamd generated. Step 5: Complete the backbone termination problem with ProCamd. Workshop on Product Design, NTUA, March 2011 82 Other chemical product design problems: 3f Problem Statement-3 (Generate backbone with ProCamd & terminate manually with ProPred) ProCamd generates 74 terminated structures out of which compound d 1 is i Ibuprofen Ib f Step 4: Launch ProCamd from ProPred from the “tools” menu A new ProCamd application is opened. ProPred sends back the same backbone structure that ProCamd generated. Step 5: Complete the backbone termination problem with ProCamd. Workshop on Product Design, NTUA, March 2011 83 Other chemical product design problems: 4a Problem Statement-4 (Design large molecules): Design a large molecule having the following p p properties, Mw > 300 Tb > 400 K See page 61 of tutorial document Tm > 300 K Solve the problem with ProCamd and then switch to ProPred and further investigate the properties of the l large molecule, l l including i l di further f h increase i off the h size i of the molecule. Only the “general problem control” and the “non-temperature non-temperature depd props” props need to be specified. Workshop on Product Design, NTUA, March 2011 84 Other chemical product design problems: 4b Problem Statement-4 Design a large molecule having the following p p properties, Mw > 300; Tb > 400 K; Tm > 300 K Solve the problem with ProCamd and then switch to ProPred. Repeat the problem for acyclic compounds d and d cyclic li compounds Workshop on Product Design, NTUA, March 2011 85 Exercises: summary • Solve example problem 1 (find best AI) • Solve example problem 2 (backbone termination) • Solve example problem 3 (backbone generation & termination) • Solve example problem 4 (generate large molecules, pass to ProPred and then analyze and further develop ) Workshop on Product Design, NTUA, March 2011 86 Summary • For computer aided product-process design, models are necessaryy – If appropriate models are available, almost all problems can be solved • Solution of different product-process design problems need a good understanding of the problems so that the available tools can be applied correctly and efficiently • Problems related to low low-value value products are easier to solve than the high-value products – The limiting factors for these problems are availability of data and appropriate models Workshop on Product Design, NTUA, March 2011 87