Break­ing down si­los with a new quan­ti­ta­tive ap­proach to drug de­vel­op­ment

Bio­phar­ma­ceu­ti­cal re­search and de­vel­op­ment (R&D) is a la­bo­ri­ous, re­source-in­ten­sive process re­quir­ing time and a sig­nif­i­cant cap­i­tal ex­pen­di­ture, with­out any guar­an­tee that a drug can­di­date will ever make it to reg­u­la­to­ry re­view and ap­proval, let alone in­to the hands of pa­tients who might ben­e­fit from it. Yet, de­spite the in­ef­fi­cien­cies of cur­rent drug de­vel­op­ment, change to speed up the drug de­vel­op­ment process has been slow to come.

 Key points

• The typ­i­cal drug de­vel­op­ment process in­cludes dis­crete steps to get to ap­proval, where quan­ti­ta­tive meth­ods are used but are not cen­tral to the de­ci­sion-mak­ing process.
• Quan­ti­ta­tive an­a­lyt­ics and da­ta can change the way drug de­vel­op­ment is done, from dis­cov­ery to the re­al-world set­ting.
• Break­ing down si­los across R&D groups to com­bine all avail­able da­ta and an­a­lyt­ics al­lows for crit­i­cal de­ci­sions to in­form drug de­vel­op­ment.

Ac­cord­ing to De­loitte’s 14^th An­nu­al Phar­ma­ceu­ti­cal In­no­va­tion Re­port, among the top 20 glob­al phar­ma­ceu­ti­cal com­pa­nies alone, they col­lec­tive­ly spent $145 bil­lion on R&D in 2023.¹ Mea­sured against an an­nu­al av­er­age of 53 new drug ap­provals by the U.S. Food and Drug Ad­min­is­tra­tion over the last five years,² R&D is a cum­ber­some, ex­pen­sive, risky process, and it is clear that a new ap­proach to drug de­vel­op­ment, one that iden­ti­fies and ex­ploits ef­fi­cien­cies, is sore­ly need­ed.

Break­ing away from the “typ­i­cal” way of do­ing things

The nor­mal drug de­vel­op­ment process, typ­i­cal­ly seen more as a qual­i­ta­tive process, in­volves de­sign­ing and ex­e­cut­ing dis­crete steps that progress a project for­ward. While some com­bi­na­tion of math­e­mat­ics, com­put­er pro­gram­ming, ar­ti­fi­cial in­tel­li­gence (AI), pre­dic­tive mod­el­ing, mod­ern clin­i­cal tri­al de­sign, re­al-world ev­i­dence and ge­nom­ic and mol­e­cule da­ta may play a role in dri­ving the re­search for­ward, we rarely can un­lock their full po­ten­tial be­cause we are look­ing too hard at the next ex­per­i­ment or tri­al that we need to run and not ful­ly ex­ploit­ing all of the da­ta we have avail­able to make the right next de­ci­sions.

At Jazz Phar­ma­ceu­ti­cals, we take a dif­fer­ent ap­proach to drug de­vel­op­ment, one where quan­ti­ta­tive da­ta helps in­form every as­pect of the clin­i­cal de­vel­op­ment process. It starts in dis­cov­ery where we choose which mol­e­cule to pur­sue. It im­pacts the pre­clin­i­cal phase when we are look­ing at the ge­nom­ic pro­file of a mol­e­cule. Then it moves for­ward to how we pre­dict what the mol­e­cule does to the body and what the body does to the mol­e­cule. It im­pacts how we de­sign our clin­i­cal tri­als to de­ter­mine if our in­ves­ti­ga­tion­al med­i­cines are safe and ef­fec­tive, and it looks at how we an­a­lyze the da­ta in the re­al-world set­ting.

How­ev­er, this does not mean that we have com­plete­ly aban­doned the qual­i­ta­tive ap­proach. As head of da­ta sci­ence, my team and I are laser fo­cused on iden­ti­fy­ing the in­ter­sec­tion of qual­i­ta­tive and quan­ti­ta­tive re­sults. As a clin­i­cal sta­tis­ti­cian by train­ing, I know the im­por­tance of bring­ing to­geth­er cross-func­tion­al teams to strate­gi­cal­ly guide drug de­vel­op­ment process­es from start to fin­ish based on da­ta.

In that re­gard, I do think our in­dus­try has done a good job of bal­anc­ing the use of these meth­ods, on­ly in more of a siloed way. Typ­i­cal­ly, a com­pa­ny might use a cer­tain method in bioin­for­mat­ics, a method in clin­i­cal tri­al de­sign, a method in re­al-world ev­i­dence. What I feel we as an in­dus­try have not done a good job of is mak­ing those meth­ods cen­tral to drug de­vel­op­ment and con­nect­ing them across dif­fer­ent dis­ci­plines.

What I be­lieve makes the Jazz ap­proach to da­ta sci­ence with­in R&D unique is that we have cen­tral­ized many of the da­ta sci­ence func­tions in­to one de­part­ment, rec­og­niz­ing the need to break down these si­los to ful­ly em­brace the val­ue of da­ta sci­ence across R&D or­ga­ni­za­tions. As the use of da­ta con­tin­ues to evolve, as well as the po­ten­tial of mul­ti­ple sources of da­ta to in­form de­vel­op­ment and reg­u­la­to­ry path­ways – in­clud­ing re­al-world ev­i­dence, clin­i­cal tri­als and mod­el-in­formed drug de­vel­op­ment – the need for in­te­grat­ed da­ta sci­ence is even more im­por­tant to­day. How­ev­er, I find that these three dif­fer­ent av­enues present the chal­lenge of be­ing siloed from one an­oth­er.

Em­brac­ing an in­te­grat­ed ap­proach to dri­ve de­ci­sion mak­ing

Jazz takes a unique ap­proach to da­ta sci­ence, with in­ter­con­nect­ed da­ta sci­en­tist teams that have broad ex­po­sure to our quan­ti­ta­tive dis­ci­plines that can ap­proach the three av­enues to­geth­er, rather than apart. We have cre­at­ed a cul­ture of Quan­ti­ta­tive Drug De­vel­op­ment, where the quan­ti­ta­tive drug rev­o­lu­tion comes in – bring­ing di­verse per­spec­tives and da­ta to all sce­nar­ios we might face. Quan­ti­ta­tive drug de­vel­op­ment in­volves the in­ten­tion­al, fo­cused and in­te­grat­ed use of all avail­able da­ta and an­a­lyt­ics to in­form and make key de­ci­sions about med­i­cines in de­vel­op­ment.

This strat­e­gy al­lows us to de­liv­er a more tar­get­ed ap­proach to reach the right pa­tients at the right time with ther­a­pies that will have the great­est im­pact across our pipeline, in­clud­ing ear­ly-stage de­ci­sion mak­ing, strate­gic ways to get a prod­uct ap­proved and ev­i­dence-based de­ci­sions to con­tin­ue the suc­cess of the prod­uct once it’s in re­al-world sce­nar­ios.

The true in­no­va­tion in chang­ing drug de­vel­op­ment is go­ing to come through the in­te­grat­ed use of da­ta across the R&D life­cy­cle. On­ly by al­low­ing da­ta sci­en­tists across the or­ga­ni­za­tion to col­lab­o­rate will we tru­ly make the best de­ci­sions for our pa­tients and speed up get­ting med­i­cines to pa­tients in need. By lever­ag­ing all op­tions with­in com­plex clin­i­cal tri­al de­sign, mod­el-in­formed drug de­vel­op­ment and re­al-world ev­i­dence, we can choose the best path to push a prod­uct for­ward or cre­ative­ly com­bine dif­fer­ent op­tions to tru­ly in­no­vate and get prod­ucts to mar­ket in ways no one has done be­fore. Ma­chine learn­ing and AI can aid us at every step in this process, help­ing us to make these key de­ci­sions and even gen­er­ate new ideas we hadn’t pre­vi­ous­ly con­sid­ered.

By re­struc­tur­ing this ap­proach, com­pa­nies can be­come bet­ter quan­ti­ta­tive drug de­vel­op­ers be­cause they will have teams – in­clud­ing bioin­for­mat­ics, sta­tis­tics, re­al-world ev­i­dence, epi­demi­ol­o­gy and sta­tis­ti­cal pro­gram­ming – col­lab­o­rat­ing across R&D di­rect­ly in­stead of be­ing rep­re­sent­ed through non-quan­ti­ta­tive stake­hold­ers. This can al­low teams to con­sid­er all the fac­tors at play to help de­vel­op in­te­grat­ed an­a­lyt­i­cal plans for clin­i­cal de­vel­op­ment that will lead to ap­pro­pri­ate so­lu­tions for pa­tients.

As the sources and us­es of da­ta con­tin­ue to rapid­ly evolve, I look for­ward to the suc­cess­es teams at Jazz will have liv­ing this mod­el, as I see this po­ten­tial­ly speed­ing up our ef­forts to be able to get these drugs to mar­ket. For the bio­phar­ma­ceu­ti­cal in­dus­try, this break­ing down of si­los holds in­cred­i­ble promise to en­sure all teams are work­ing in lock­step and can col­lec­tive­ly con­tin­ue to de­vel­op and de­liv­er in­no­v­a­tive med­i­cines to pa­tients.

Learn more about Jazz Phar­ma­ceu­ti­cals’ ap­proach to R&D here.

Ref­er­ences

¹ De­loitte. Un­leash AI’s po­ten­tial: Mea­sur­ing the re­turn from phar­ma­ceu­ti­cal in­no­va­tion – 14th edi­tion. April 2024. https://www2.de­loitte.com/con­tent/dam/De­loitte/us/Doc­u­ments/life-sci­ences-health-care/us-rd-roi-14th-edi­tion.pdf. Ac­cessed June 2024.

² de la Torre BG, Al­beri­cio F. The Phar­ma­ceu­ti­cal In­dus­try in 2023: An Analy­sis of FDA Drug Ap­provals from the Per­spec­tive of Mol­e­cules. Mol­e­cules. 2024;29(3):585. doi:10.3390/mol­e­cules29030585