ChemiCal eduCation CHIMIA 2018, 72, No. 1/2 55doi:10.2533/chimia.2018.55 Chimia 72 (2018) 55–61 © Swiss Chemical Society

*Correspondence: Prof. Dr. V. Zoeteae

E-mail: vincent.zoete@sib.swissaMolecular Modeling Group, SIB Swiss Institute ofBioinformaticsBâtiment Génopode, Quartier Sorge,CH-1015 LausannebOutreach Team, SIB Swiss Institute of BioinformaticsBâtiment Génopode, Quartier Sorge,CH-1015 LausannecSwiss-Prot Group, SIB Swiss Institute of Bioinforma-tics, CMU1 rue Michel Servet, University of Geneva,CH-1211 Geneva 4dTraining Group, SIB Swiss Institute of BioinformaticsBâtiment Génopode, Quartier Sorge,CH-1015 LausanneeDepartment of Fundamental Oncology and LudwigInstitute of Cancer ResearchUniversity of Lausanne, Route de la Corniche 9A,CH-1066 Epalinges‡A.D. and M.C.B. contributed equally to this work.

Educational Tools to Introduce Computer-Aided Drug Design to Students and to thePublic at Large

Antoine Dainaa‡, Marie-Claude Blatterbcd‡, Vivienne Baillie Gerritsenbc, and Vincent Zoeteae*

Abstract: The Drug DesignWorkshop initiative was put in place in 2015 and consists of a collection of educationaltools especially developed to introduce computer-aided drug design to the general public and students of vari-ous levels. These presentations, hands-on sessions, physical material and on-line educational tools (http://www.drug-design-workshop.ch) have been used in a variety of settings including classrooms, universities, teachertraining sessions and science fairs. The main goal is to inform an audience as broad as possible regarding theprocesses and challenges for the design, discovery and development of drugs. The present article describeswhat is presently available and the future direction for new activities currently under development.

Keywords: Drug discovery · Computer-assisted drug design · Education · General public · STEM

Antoine Daina holds a FederalDiploma of pharmacy and obtained aPhD in Pharmaceutical Sciences from theUniversity of Geneva in 2006. He spentthree years as a computational chemist atSyngentaCropProtection, in theChemistryResearch Department and three years atthe School of Pharmaceutical Sciencesin Geneva as Senior Scientist/Lecturer.

Antoine Daina joined the MolecularModeling Group at SIB, Swiss Institute ofBioinformatics in 2012 and is in charge ofdeveloping, providing and applying novelmethods for computer-assisted drug de-sign. His main research themes includechemoinformatics and machine-learningmodels for physicochemistry, pharmaco-kinetics andADME. Besides his academicteaching duties, he is involved in the devel-opment of educational tools for drug de-sign targeting high-school students and thegeneral public, in the context of the DrugDesign Workshop project.

Marie-Claude Blatter has a PhD inBiochemistry. After several years in bio-medical academic research, she is nowresponsible for the outreach activities atthe SIB Swiss Institute of Bioinformatics.She coordinates events for classes (age12–19) and for the public at large, takes anactive part in them, and is responsible fortraining high-school teachers. She is alsoinvolved in biocuration and user supportfor the UniProtKB protein knowledgebase(www.uniprot.org) and is part of academicteaching program with special expertise inthe fields of biological data & knowledge-bases.

Vivienne Baillie Gerritsen holds adegree in Biology from the University ofGeneva, and has been involved in popularscience for many years. She is a sciencewriter at SIB, and author of the ‘ProteinSpotlight’ articles (www.proteinspotlight.org).

Following an engineer degree from the‘Ecole Nationale Supérieur de Chimie deLille’, a Master in Drug Design and a PhDin organic chemistry from theUniversity ofLille in 1999, Vincent Zoete made a Post-doc in molecular modeling and drug de-sign in Prof. Karplus’ group in Strasbourg,and subsequently in Prof. Meuwly’s groupin Basel. He then joined the SIB SwissInstitute of Bioinformatics in 2004 wherehe was appointed as Associate GroupLeader of the Molecular Modeling Group

56 CHIMIA 2018, 72, No. 1/2 ChemiCal eduCation

CADD (‘computers’) supports the evalu-ation of key properties for a molecule tobecome an actual drug, e.g. affinity forthe target, fate in the organism or possibleside-effects.

2.1 3D-printed Protein and DrugsAlthough executed by computers, we

felt that it was important for the audienceto realize that CADDmethods are based ontangible physical concepts. For instance,the fundamental notion ofmolecular recog-nition of a small drug molecule by a targetprotein can be modeled with rules relatedto classical mechanics (so-called molecu-lar mechanics) including various potentialsfor internal and non-bonded energy terms(like Lennard-Jones or Coulomb poten-tials). The evaluation of how a small mol-ecule binds to the macromolecular protein,and how strongly, are performed in silicoby docking engines. These computer pro-grams are able to predict the most probableconformation, orientation and position ofthe small molecule at the surface of theprotein by optimizing the interactions be-tween both molecular partners at the atom-ic level. The underlying ‘lock-and-key’concept of docking is intuitive.As a meansof introduction to the general public andstudents, a model of the protein (cyclooxy-genase-1, COX1) targeted by nonsteroidal

sions we have had stressed the need for agreater perception of 1) the drug develop-ment workflow and how computers dealwith it (mainly among the younger par-ticipants); and 2) current challenges likepersonalized medicine or drug resistance(mainly for the well-informed public).This fostered the creation of additionalmaterial to broaden even further our tar-geted audience.

2. Drug Design Workshop – WhatExists

Different educational materials – gath-ered under the general initiative ‘DrugDesign Workshop’ – were conceived in aflexible way so as to be useable in a largevariety of settings while targeting a diverseaudience (pupils, high-school and univer-sity students as well as the general public).Developed activities (Table 1) share thesame pedagogical objectives: i) to informthat most drugs are small synthetic mol-ecules which interact with a protein in thebody to trigger a therapeutic effect; ii) toreveal how time-consuming, complex andexpensive drug discovery and developmentprocesses are; iii) to show that the processof discovering new drugs is a collective,interdisciplinary effort; iv) to explain how

in 2011. He is also Assistant Professor inthe Department of Fundamental Oncologyat the University of Lausanne since 2017,and the Scientific Director of the ProteinModeling Facility of the same university.Together with Prof. Michielin, VincentZoete is in charge of the SwissDrugDesignproject of the SIB, which consists ofa collection of freely accessible webtools for computer-aided drug design, in-cluding SwissDock.ch, SwissParam.ch,SwissSidechain.ch, SwissTargetPrediction.ch, SwissBioisostere.ch, SwissSimilarity.ch and SwissADME.ch. He is also activein the field of education, notably in com-puter-aided drug design with the Drug-Design-workshop.ch project.

1. Introduction – Motivations

The chemical sciences have shown re-markable benefit for society. In particular,over the past century life expectancy andquality has been favorably impacted by thediscovery, development and production ofnew pharmaceuticals.[1]

Furthermore, ‘drugs’ as a whole is apopular theme among school students,and drug discovery is itself an appreciatedsubject at university level. The success of anew treatment enables high visibility of thediscipline for the public, which is informedby the generalist media about general con-cepts and global cost of ‘making’ drugs.[2]However, the concrete challenges of drugdiscovery, and the research and technologi-cal tools that are available to achieve them,remain largely unknown to the generalpublic and students. Here we focus on theapproaches and methods which make useof computing resources – algorithms, data-bases and 3D-visualization – and are cap-tured by the term Computer-Aided DrugDesign (CADD). CADD aims at drawingrational hypotheses by performing chemi-cal/structural analyses in order to generatedecision-making criteria which support theexecution of drug-discovery processes –the final aim being to improve efficiencyin providing pharmaceutical agents for thepatient.[3–5]

In this article, we describe our DrugDesign Workshop, which consists of acollection of educational tools for CADDthat are developed and maintained by us.These presentations, hands-on sessions,and on-line educational tools (http://www.drug-design-workshop.ch) are conductedin a variety of settings – from high-schoolclasses and specialized educator trainingsto science fairs – with great success. Asan example, over 1’200 high-school stu-dents have taken part in our workshopsreporting an overall high satisfaction level(5.0 out of 6.0, on average). The positivefeedbacks and many constructive discus-

Table 1. List of Drug Design Workshop activities described in the present article.

Activity Short description Public

Educationalwebsite

www.drug-design-workshop.ch includesthree computer-assisted workshops, wherethe objective is to design the ‘best’ possiblemolecule to treat either inflammation,melanoma or other cancers.

12 years andolder

3D-printedprotein anddrugs

The COX protein and several anti-inflammatory drugs, printed at the same scale,allow for an intuitive hands-on illustrationof molecular docking and intermolecularrecognition.

8 years andolder

Wooden puzzle Young children can understand that it ispossible to design diverse molecules able tofit into a protein cavity.

5 – 8 years

Deduction cardgame

This game illustrates the multi-objectivenature of drug discovery by searching themolecule with all optimal properties (i.e.non-toxic, water-soluble, well-absorbed,stable, efficient in humans and selective to thetarget).

8 years andolder

Molecularfingerprintsexercise

The participant can reproduce with pen-and-paper the process applied by a computer tobehold molecular structures and quantify thesimilarity between them. It illustrates the‘similarity principle’ in drug design.

14 years andolder

ChemiCal eduCation CHIMIA 2018, 72, No. 1/2 57

rect access to a page including a 3D sessionof the pre-calculated docking pose. Thissession is interactive so that the user canzoom, rotate and translate the system withthe mouse, to analyze visually the predictedbinding mode.

Briefly, as shown in Fig. 2 for the COXworkshop, the input page includes picturesof the 3D structure of target proteins onthe left, and boxes including 2D chemicalstructures of representative drugs on theright. A simple drag-and-drop of a givenchemical structure on one protein gives di-

anti-inflammatory (NSAIDs) compoundswas 3D-printed, as were the active ingre-dients of well-known anti-inflammatorydrugs (e.g. ibuprofen, diclofenac, nimesu-lide). At first, the 3D-printed models areuseful for beholding the relative sizes of themacromolecular protein and of the smallmolecule drugs. Manual docking can thenbe performed by opening the printed COX1structure so that the binding site becomesaccessible, and by positioning the printeddrug molecule inside (such as ibuprofen asin Fig. 1). This approach is very approxi-mate from a physics point of view because itonly accounts for shape recognition and notfor specific forces (like hydrophobic forcesor those driven by hydrogen-bonding) normolecular flexibility for instance. But webelieve that this manual exercise helps toreveal clearly both the challenges of dock-ing and the need for automation throughcomputer algorithms. Such algorithms areable to increase dramatically the number ofmolecular structures which need to be han-dled while decreasing the extent of physicalapproximations.

2.2 Web-Based Educational CADDA fully integrated Web interface was

created to initiate beginners to the use ofcomputers for designing molecules andevaluating their potentiality to become anactual drug. The website is freely availableat www.drug-design-workshop.ch for theEnglish version or at www.atelier-drug-design.ch for the French version (transla-tion in German is in progress), and can beconsidered as the cornerstone of the DrugDesignWorkshop. It involves introductorymovies, documents, help pages and threeworkshops regarding pertinent therapeutictargets. Three proteins relevant to drug de-sign were chosen: (i) both isoforms of cy-clooxygenase (COX1 and COX2) with theaim of introducing the concept of specific-ity for a protein with regards to therapeuticversus unwanted effects of NSAIDs; (ii)B-Raf kinase for the notion of somaticmutation and targeted cancer therapy withgame-changing drugs; like vemurafenib;and (iii) indoleamine 2,3-dioxygenase 1(IDO1) as an example of current cancerimmunotherapy targets involving researchdrug-candidates molecules. The biologi-cal context and usage of the workshopsare extensively provided in dedicated Webpages and in a previous article.[6] The maingoal of these on-line workshops is to letthe user enter the iterative cycle of design-ing and optimizing a molecule to make it astrong ligand for the protein target. This isachieved by hiding the technical complex-ity behind a simple and user-friendly inter-face, and by using enough approximationsto keep the calculation time short (abouttwo minutes or less) for the process to betruly interactive.

Fig. 1. The 3D model of cyclooxygenase-1 (COX1) and ibuprofen. (A) The computer model ob-tained from the crystal structure resolved by X-ray diffraction (pdb entry 1EQG) showing COX1(molecular surface in white), the cutting plane and ibuprofen (molecular surface in beige) accom-modated in the binding site. (B) The printed model employed in classroom for manual dockingof ibuprofen in the opened COX1.

Fig. 2. An example of the Drug Design Workshop website input page, where technical complex-ity is hidden behind a user-friendly interface to allow simple drag-and-drop docking and easysketching of molecules. Adapted from ref. [6].

58 CHIMIA 2018, 72, No. 1/2 ChemiCal eduCation

For more advanced participants, thepossibility to access two professional re-search tools is given. By clicking on oneof the two buttons at the bottom of theresult page, the user’s molecule is sub-mitted either to SwissTargetPrediction[8]to estimate its most probable proteintargets or to SwissADME[9] for the cal-culation of physicochemical, pharmaco-kinetic and drug-likeness properties. Theuser can thus get a realistic ‘feel’ of thedifference between research and educa-tional tools. More important maybe iswhat the participant learns from theseexpert methods: i.e. that for a moleculeto become a drug, it has not only to berecognized by its protein targets for po-tency, but it must also fulfil a number ofcriteria such as being non-toxic, show-ing few side-effects, baring the optimalproperties to reach the protein target inthe body, being quite rapidly eliminated,being synthesizable and stable, just tocite a few.With guidance from experts orwell-trained educators, the central notionof CADD, which consists in a multipleobjective iterative cyclic process, can beclearly understood.

This link to specialized methodologiesthat are systematically used in researchsettings was the first answer to demandsthat were formulated on feedbacks, andwhich asked for the workshop to be ex-tended to a more informed audience.What is more, the biological context ofthe B-Raf workshop was enriched so asto touch the theme of personalized medi-cine. Focus is given on the importance ofdetermining whether the melanoma cellsof a given patient bears a specific mutation(V600E) or not. By sequencing its BRAFgene (Uniprot ID: P15056), the patient isdefined eligible for treatment with drugsthat inhibit this specific mutant, such asvemurafenib. In the case of melanomacells that do not carry the V600E muta-tion of B-Raf, vemurafenib proved tobe deleterious because it favored tumorgrowth.[10] The blocking of melanomacell proliferation by this medium consti-tutes a recent success story in the realm ofdrug design in cancer targeted therapy andpersonalized medicine. A movie exempli-fying these aspects at molecular level isavailable on the website. Other workshopsin the context of individualized therapyand drug resistance shall be implementedin the near future.

3. Broadening the Audienceto a Younger Public

The other major demands were to ex-tend the targeted audience to youngerparticipants, either in the classroom or inscientific events. Following this, various

formed on a remote server, computationtakes about 20 seconds to 2minutes depend-ing mainly on the size of binding site andligand. Moreover, depending on the load ofthe server, the calculation can be queued.A panel informs the user on the progressof its run. In a standard setup, 15 users canperform basic steps on the website simulta-neously without affecting interactivity.

Upon completion, one click displaysthe binding mode as an interactive 3Dsession in the result page, which also pro-vides an evaluation of the binding strength(Fig. 3). For the sake of clarity, this scoreis the opposite of the binding free energyas predicted byAutodockVina.[7] This per-ception of ‘the larger the score, the betterthe ligand’ and the relative scale for com-paring with reference drugs is intuitiveenough to strongly motivate the users todesign ‘better’ molecules, and thus to entereffortlessly into the characteristic iterativeoptimization cycle inherent to CADD.

The user is invited to design his/her owntentative drug by clicking on the ‘Designyour own molecule’ box of the input page.A molecular sketcher opens (MarvinJSversion 6.1, 2013, www.chemaxon.com) todraw a chemical structure. For less experi-enced users, the sketcher can be filled withone of the pre-selected drugs, by clickingthe down red arrow next to the structure.Any modifications can be then applied byusing the sketcher toolbars. In our experi-ence, even students without knowledge inorganic chemistry are able to draw correctmolecular structures with the help of thesketcher facilities.

By clicking the ‘Done’ button, the mol-ecule appears in the corresponding box, andis available for docking with the same drag-and-drop procedure as explained above.

All docking technical steps, i.e. prepa-ration of ligand and protein as well as thedocking calculation by AutoDock Vina,[7]are transparent to the user. Although per-

Fig. 3. Example of a result page for a molecule designed by a user of the Drug Design Workshopwebsite and docked into COX2. The upper part is dedicated to the 3D interactive session tovisualize the predicted binding mode; the middle part allows the comparison of the predictedpotency (score) of the ‘virtual’ molecule with ‘real’ existing drugs; on the bottom of the page aretwo buttons that enable easy submission to research expert tools SwissTargetPrediction andSwissADME for further assessment of the potential of the user’s molecule to become a drug.Adapted from ref. [6].

ChemiCal eduCation CHIMIA 2018, 72, No. 1/2 59

activities were developed to depict and ex-emplify key notions of CADD especiallyfor youngsters.

3.1 Wooden PuzzleThe activities regarding molecular rec-

ognition and iterative cyclic optimizationdescribed above, i.e. the 3D-printed mod-els for the manual docking NSAIDs (Fig.1) and the online workshop for virtualdocking (Figs 2 and 3) have proven to beexcellent tools for introducing importantconcepts of CADD. However, they turnedout to be too advanced for the younger par-ticipants. Therefore, we designed a simplepuzzle in wood, which enables childrenfrom age 5 onwards to build and docksmall molecules into a 2D protein, by fill-ing the binding site cavities with puzzlepieces representing molecular fragments(Fig. 4). This tool introduces effortlesslythe notion of shape complementarity andscaffold hopping since different pieces canfill the same pocket. Younger children areable to understand that diverse moleculescan achieve the ‘lock-and-key’ objective,while more advanced participants can cal-culate how many well-fitting moleculescan be generated.

3.2 Deduction Card GameFinding a molecule with a shape that

fits into the protein binding site is certain-ly needed but not sufficient to propose adrug candidate. Other aspects such as themolecule’s fate in the organism, possibleside effects and toxicity must be investi-gated. To illustrate this important aspectof CADD, we have developed a deduc-tion game, whose goal is to find among sixcards with one molecule on each, the bestdrug candidate (Fig. 5). The best candidateis the molecule that displays all optimalproperties, i.e. non-toxic, water-soluble,well-absorbed, stable, efficient in humansand selective to the target. All other fivepossible cards showmolecules with at leastone sub-optimal property. Each property islinked to a specific fragment in the mol-ecule, and this link has to be deduced froma pool of cards displaying molecules withsub-optimal properties (for an example,refer to Fig. 5A). The number of cardscan vary so as to adapt the difficulty to thelevel of the participants (from the age of 8).Once the participant has attributed a givensub-optimal property to a molecular frag-ment, he/she can identify, among the sixmolecules mentioned above, the one thatdoes not show any of the problematic frag-ments. Of note, participants do not need tomaster concepts in organic chemistry tobe able to do this exercise: understandingis straightforward when considering mol-ecules as images with a common part (thescaffold) and differences (the molecularfragments). The tutors can then introduce

the concept of chemical structures if re-quired, or useful.

3.3 Molecular Fingerprints ExerciseApart from methodologies using infor-

mation about the structure of the proteintargets (‘structure-based’), other CADDapproaches depend on the knowledge ofsmallmoleculeswithdesiredactivities.Therational of such ‘ligand-based’ approachesis founded on the similarity principle,which assumes that similar molecules areprone to exhibit similar biological activi-ties.[11,12] This central concept can, for in-stance, be used to perform virtual screen-ing: i.e. looking into a library of (commer-cially) available molecules that are similarto an active one, andwhich should be testedexperimentally in priority. Computers areneeded to be able to rapidly quantify mo-lecular similarities between the millions ofmolecules included in chemical librariesnowadays. The way a computer performsthis task for a couple of molecules pro-vides an excellent opportunity for STEM(Science, Technology, Engineering andMathematics) teaching,[13] by making a di-rect link between mathematics, informat-ics, chemistry and drug design. We havedesigned a pen-and-paper exercise so thatparticipants can actively experience howto translate an intuitive qualitative estima-tion of molecular similarity – like a hu-

man can do – into a quantitative evaluationby simple maths and an algorithm appli-cable through a computer program. Withthis in mind, we used a technique called‘molecular fingerprints’ that translates achemical structure into a bit string of 0and 1, which is the actual regular object ofcomputer calculations.[14,15] A simplifiedprotocol is given to the participant to cre-ate fingerprints and manipulate them forcalculating molecular similarity ‘by hand’(Fig. 6). For a series of drug molecules,with the help of a transparent small ruleron which chemical fragments are drawn,participants note the presence or absenceof fragments by writing in the related bitof an empty vector either ‘1’ (fragment isfound in the molecule), or ‘0’ (not found).The so completed vectors of eachmoleculeare then compared two-by-two by calculat-ing the Tanimoto coefficient (TC).[16]TC isobtained by dividing the number of times‘1’ is found in the same column of bothvectors by the number of times ‘1’ is foundin at least one of the two vectors. The no-tion of vector does not necessarily need tobe introduced, since it simply (and natu-rally) translates into a table in this exercise.Finally, by confronting TC between knowntherapeutic classes of molecules (here anti-inflammatory versus anti-cancer drugs) theapplication of the similarity principle be-comes concrete.

Fig. 4. Wooden puzzle for young children to perform manual docking (A) first outline; and (B) well-advanced prototypes tested in a science fair setting.

Fig. 5. Deduction card game dedicated for multi-objective optimization. (A) Design of a card usedfor deduction: in this example, the molecule, three sub-optimal properties, i.e. toxicity (blackskull), instability (green Pisa tower) and non-specificity (purple organs), are to be linked with thethree fragments of the molecule; and (B) prototypes of cards tested during a science fair.

60 CHIMIA 2018, 72, No. 1/2 ChemiCal eduCation

to school teachers, and even Bachelor orMaster students. The general public isreached in science fairs or other events.

Since 2015, more than 1’200 highschool students have attended the work-shops and, with our guidance, have de-signed and evaluated molecules for theirpotential to become a drug. Note that ourwebsite has also been designed simpleenough so that oneperson canuse itwithoutthe need of tutor guidance. Since 2015, ithas received around 10’000 unique visitors

4. Outcome and Outlook

The Drug Design Workshop activi-ties are all independent from one anotherand have different pedagogical objectives.However, they can also be presented in alogical manner. Participants can do one orseveral activities depending on their avail-ability and level of understanding. Thisallows us to target a wide audience, fromthe youngest (aged 5–12), secondary I(12–15) and secondary II (15–19) students

For full practicality, reusable physi-cal supports including all information (atthe moment available only in French as inFig. 6C) were produced for sciencefairs and classrooms. In our experience,the general public and children from14 years old are able to do this exerciseand understand the concrete applicationof mathematics and engineering to thefield of chemistry. An online version ofthe exercise is planned to be added to ourDrug Design website.

Quels fragments se trouventdans ces molécules?

À l’aide de la réglette, retrouve les fragmentsprésents dans chaque molécule.Mets un 1 dans les cases correspondantessi tu les trouves et 0 sinon. Exemple :

1 1 1 0 1 0 0 1

Diclofenac

Lumiracoxib

Erlotinib

Gefitinib

A

C

Calcul de similaritéCœfficient de Tanimoto

Quelles sont les paires de molécules les plus similaires?

A

B

0 ≤ T ≤ 1 T = 0: molécules totalement différentesT = 1: molécules identiques

nombre de cases où l’on trouve1 à la fois chez A et chez B

nombre de cases où l’on trouve0 chez A et 1 chez B

nombre de cases où l’on trouve1 chez A et 0 chez B

Exemple

Formule

Compare les 4 molécules en remplissant le tableau ci-dessousavec les cœfficients de Tanimoto calculés :

Diclofenac

Diclofenac

Lumiracoxib

Lumiracoxib

Erlotinib

Erlotinib

Gefitinib

Gefitinib

1

1

1

1

1

0

1

1

1

1

0

1

1

1

0

0

0

0

0

1

B

Fig. 6. STEM exercise on Molecular Fingerprints. (A) Design of the tablet’s first page, which allows to complete the binary bit strings as a computerwould translate molecular structures; (B) Design of the tablet’s second page, giving all information on how to quantify chemical similarity betweenmolecules; and (C) young students employing the reusable supports and the small transparent rulers to calculate the molecular similarity betweenpairs of drugs.

ChemiCal eduCation CHIMIA 2018, 72, No. 1/2 61

acknowledge for the academic license agree-ment. Molecular docking is performed byAutodockVina and the 3D session embeddedwithin the web pages are JSmol applets.

Received: September 29, 2017

[1] F. R. Lichtenberg,Health Policy andTechnology2014, 3, 36.

[2] T. R. Helgren, T. J. Hagen, J. Chem. Educ.2017, 94, 345.

[3] V. Zoete, A. Grosdidier, O. Michielin, J. Cell.Mol. Med. 2009, 13, 238.

[4] G. Schneider, Chimia 2012, 66, 120.[5] W. L. Jorgensen, Acc. Chem. Res. 2009, 42,

724.[6] A. Daina, M.-C. Blatter, V. Baillie Gerritsen, P.

M. Palagi, D. Marek, I. Xenarios, T. Schwede,O. Michielin, V. Zoete, J. Chem. Educ. 2017,94, 335.

[7] O. Trott, A. J. Olson, J. Comput. Chem. 2010,31, 455.

[8] D. Gfeller, A. Grosdidier, M. Wirth, A. Daina,O. Michielin, V. Zoete, Nucleic Acids Res.2014, 42, W32.

[9] A. Daina, O. Michielin, V. Zoete, Sci. Rep.2017, 7, 42717.

[10] G. Bollag, J. Tsai, J. Zhang, C. Zhang, P.Ibrahim, K. Nolop, P. Hirth, Nat. Rev. DrugDiscov. 2012, 11, 873.

[11] Y. C. Martin, J. L. Kofron, L. M. Traphagen, J.Med. Chem. 2002, 45, 4350.

[12] A. Bender, R. C. Glen, Org. Biomol. Chem.2004, 2, 3204.

[13] S. Freeman, S. L. Eddy, M. McDonough, M. K.Smith, N. Okoroafor, H. Jordt, M. P.Wenderoth,Proc. Natl. Acad. Sci. U.S.A. 2014, 111, 8410.

[14] I. Muegge, P. Mukherjee, Exp. Opin. DrugDiscov. 2016, 11, 137.

[15] A. Cereto-Massaguà, M. J. Ojeda, C. Valls, M.Mulero, S. Garcia-Vallvé, G. Pujadas, Methods2015, 71, 58.

[16] P. Willett, J. M. Barnard, G. M. Downs, J.Chem. Inf. Model. 1998, 38, 983.

latest discoveries in biomedical researchand Life Sciences. In that respect, we con-tinue to develop additional activities inthe context of the Drug Design Workshopto introduce the most recent concepts intherapeutics, including personalized medi-cine and drug resistance. Beside informa-tion of the public at large, one objective isto stimulate and frame scientific curiosityand to help students to discover new areasof research, with the hope of facilitatingtheir career choice.

Associated contentShort movies and documents introducing

the notions mentioned in the present article areavailable under the CC-BY-ND-NC license onthe Drug Design Workshop web site at http://www.drug-design-workshop.ch for the Englishversion and http://www.atelier-drug-design.chfor the French version.

AcknowledgmentsThis work was supported by SIB, Swiss

Institute of Bioinformatics and by the SwissNational Science Foundation through Agoragrants CRAGP3_151515 (to Prof. OlivierMichielin and V.Z.) and CRAGP3_171653(to V.Z. and A.D.). The author would like tothank colleagues at the SIB, Swiss Instituteof Bioinformatics, who have been involved inconducting or elaborating parts of the DrugDesign Workshop, in particular Drs. DianaMarek and Patricia Palagi. Profs. OlivierMichielin and Ioannis Xenarios are also grate-fully acknowledged. Noémi Savary, studio KO,Yverdon-les-Bains, Switzerland was extensive-ly involved in the design and making of moviesand pen-and-paper supports. Caroline Emmelotat Decologic, Vevey, Switzerland producedthe wooden puzzle prototypes. ChemAxon is

who have opened about 15’000 sessions.27% of the session were from Switzerland,19% from the United States and the rest isspread worldwide.

According to the comments we havereceived, the participants appreciated theconcreteness and realism of the activities.It allowed them to use bioinformatics andprofessional tools in a real scientific pro-cess. Participants also liked the given over-view of how scientists work nowadays.They appreciated observing and experi-encing the various steps and parametersrequired for the design of a new drug. Lastbut not least, our activities were appreci-ated because they differ from the usualclassroom activities, mainly thanks to theirinterdisciplinary nature (chemistry, biol-ogy, mathematics, informatics and medi-cine). This highlights the usefulness of theDrug Design Workshop for STEM teach-ing which remains challenging, especiallyregarding high-school teacher training.

Our workshops have also proved to bepractical in introducing drug design con-cepts to the general public in scientific pop-ular events.About 2’500 visitors, includingpupils from the age of 8, were introduced toCADD during several science fairs and uni-versity open house days. They appreciatedbeing offered the opportunity to discoveran area of science related to drugs andmed-icine that concerns each and every one, andwhich is often the subject of debate. Theproposed activities rendered opportunitiesto discuss drug discovery and developmentpipeline, duration, costs and challenges.

We consider that it is of paramount im-portance to stay as close as possible to the