Speed­ing up drug dis­cov­ery with Ar­ti­fi­cial In­tel­li­gence

Andrew Hop­kins CEO, Ex­sci­en­tia

BioSpectrum (Asia) - - Bio Content - For more in­for­ma­tion visit web­site www.biospec­tru­ma­sia.com

It’s es­ti­mated that, on av­er­age, to bring one new drug to the mar­ket can take 1,000 peo­ple, 12-15 years, and up to $1.6 bil­lion. Phar­ma­ceu­ti­cal com­pa­nies are grap­pling with these scan­dalous costs and con­stantly in search for in­ef­fi­cien­cies in the drug dis­cov­ery process, and hope to fas­ten the process by fo­cus­ing on key el­e­ments, like man­age­ment mod­els, new tech­nolo­gies and creative study ap­proaches. With many new viruses and bac­te­ria daunt­ing the globe there is an in­creased ne­ces­sity to in­no­vate and de­velop new drugs and safe­guard pub­lic health. Also with patent cliff loom­ing large, pharma com­pa­nies need to evolve their R&D ef­forts to en­sure that the core of their busi­ness keeps pace with the changes. With in­creas­ing com­pe­ti­tion, ex­perts are look­ing at newer meth­ods to speed up and in­crease ac­cu­racy rates. One such ap­proach is us­ing Ar­ti­fi­cial in­tel­li­gence in drug dis­cov­ery.

Ex­sci­en­tia also signed a 250 mn Eu­ros deal with Sanofi in May. As part of this agree­ment, Ex­sci­en­tia will be re­spon­si­ble for all com­pound de­sign, whilst chem­istry syn­the­sis will be de­liv­ered by Sanofi. With its unique AI plat­form, Ex­sci­en­tia is de­liv­er­ing a pipe­line of ef­fi­ca­cious, bis­pe­cific small mol­e­cules, as well as highly se­lec­tive sin­gle tar­get can­di­dates, for mul­ti­ple in­di­ca­tions. In an in­ter­ac­tion with Andrew Hop­kins, CEO, Ex­sci­en­tia.

Ex­cerpts of the in­ter­view

What is ar­ti­fi­cial in­tel­li­gence (AI) and how does it speed up drug dis­cov­ery process?

AI, if used cor­rectly, is a new ap­proach to drug dis­cov­ery, that uses com­puter al­go­rithms to search cre­ate orig­i­nal so­lu­tions to the ques­tions a drug dis­cov­ery sci­en­tist might ask when de­sign­ing a new medicine. These sys­tems are uniquely able to learn from ex­ist­ing data re­sources, much in the way that a hu­man would learn and then ap­ply the knowl­edge gained on a new pro­ject. How­ever the amount of data now avail­able is so vast that it is be­yond a hu­man’s ca­pa­bil­ity. For AI fo­cused to­wards drug de­sign, where Ex­sci­en­tia fo­cuses its ex­per­tise, the typ­i­cal sources of in­for­ma­tion could be the vast re­sources of chem­i­cal struc­ture, phar­ma­col­ogy, bioas­says data as well as other sup­port­ing lit­er­a­ture and patent in­for­ma­tion. The AI al­go­rithms then ap­ply the dis­til­la­tion to the de­sign of new small mol­e­cules. Suc­cess­ful mol­e­cules will hit the de­sired tar­get whilst at the same time avoid­ing known se­lec­tiv­ity, tox­i­col­ogy or pharmoki­netic is­sues (among many other pa­ram­e­ters). Fur­ther re­fine­ment can lead to com­pletely new and op­ti­mized mol­e­cules (and IP) for ad­vanc­ing to­wards the clinic. AI is also be­ing ap­plied to many ar­eas of drug devel­op­ment. For ex­am­ple AI ap­proaches might look at bet­ter pa­tient strat­i­fi­ca­tion for clin­i­cal tri­als, thereby fit­ting the pa­tient to the treat­ment be­ing tested bet­ter, en­abling quicker re­cruit­ment and in­creas­ing the like­li­hood of get­ting the re­quired clin­i­cal re­sponse for reg­u­la­tory ap­proval.

Drug dis­cov­ery is fun­da­men­tally ex­pen­sive and time con­sum­ing. There­fore AI has the po­ten­tial - if well de­signed – to pos­i­tively im­pact a va­ri­ety of ar­eas such as those al­ready de­scribed. How­ever AI alone will not solve the prob­lem and the ap­proach needs to be driven by highly knowl­edge­able do­main ex­perts.

What are the key trends that will drive the growth of AI in drug dis­cov­ery for the next 5 years?

The need to re­duce R&D costs is a ma­jor drive. Lead op­ti­mi­sa­tion is the high­est cost per launched drug due to the num­ber of projects re­searched that never reach the mar­ket. Im­prov­ing dis­cov­ery ef­fi­ciency through high qual­ity can­di­dates that are dis­cov­ered ef­fec­tively would dra­mat­i­cally im­prove these met­rics. Ex­sci­en­tia’s deal with GSK is look­ing at this ex­act prob­lem, de­sign­ing can­di­dates in a highly pro­duc­tive man­ner.

Also, the need to re­duce health­care costs – com­bi­na­tion ther­a­pies (e.g in can­cer) look­ing to be­come pro­hib­i­tively ex­pen­sive – AI ap­proaches can be used to de­sign drugs that could mod­u­late mul­ti­ple bi­o­log­i­cal process (e.g block­ing tu­mour sig­nalling and sur­vival mech­a­nism, boost­ing im­mune re­sponse), in a cost­ef­fec­tive man­ner. Ex­sci­en­tia is look­ing to tackle ef­fi­cacy di­rectly by de­sign a breed of small mol­e­cules that we call bis­pe­cific small mol­e­cules. These are sin­gle small mol­e­cules with care­fully de­signed dual phar­ma­col­ogy. Ex­sci­en­tia projects fo­cus on I/O (with Evotec) , di­a­betes (with Sanofi) and psy­chi­atric dis­eases (Dainip­pon Su­mit­omo Pharma).

What fac­tors will play a critical role in the suc­cess of AI driven drug dis­cov­ery projects? De­liv­ery on the prom­ise.

Ex­sci­en­tia has al­ready demon­strated the de­liv­ery of a clinic-ready can­di­date from the start of a pro­ject in a quar­ter of the time and a quar­ter of the cost com­pared to tra­di­tional met­rics. Exsi­cen­tia’s goal is to im­prove this still fur­ther – 10x im­prove­ment – and then AI ap­proaches could be­come dom­i­nant.

In your opinion what are the ma­jor chal­lenges in the tra­di­tional drug dis­cov­ery process and how does AI help to ad­dress them?

Drug dis­cov­ery is fun­da­men­tally ex­pen­sive and time con­sum­ing. There­fore AI has the po­ten­tial - if well de­signed – to pos­i­tively im­pact a va­ri­ety of ar­eas such as those al­ready de­scribed. How­ever AI alone will not solve the prob­lem and the ap­proach needs to be driven by highly knowl­edge­able do­main ex­perts. At Ex­sci­en­tia we have some of the best medic­i­nal chemists working along­side the tech­nol­o­gists to de­velop ‘best of breed’ ap­proaches. Com­bin­ing hu­man ex­per­tise with AI power is cre­at­ing ‘cen­taur’ drug de­sign ca­pa­bil­i­ties – de­rived from ‘cen­taur’ chess – a term coined by chess grand­mas­ter Garry Kas­parov for the team­ing up of chess ex­perts with AI, which have the abil­ity to beat ei­ther hu­man or AI players.

What ac­cord­ing to you are the risks associated with use of AI tech­nolo­gies in drug dis­cov­ery?

In the case of our com­pany specif­i­cally we have no con­cerns as we are se­cure in our knowl­edge that we are do­main ex­perts and have care­fully de­vel­oped the ap­proaches over a 5 year pe­riod, dur­ing which it has been tested and re­fined in real-world drug dis­cov­ery sit­u­a­tions for part­ners. How­ever for the broader in­dus­try there is a risk that those with in­suf­fi­cient ex­pe­ri­ence will sud­denly ex­pect generic AI al­go­rithms to be a magic bul­let and solve com­plex tasks even in the ab­sence of do­main ex­per­tise. This is un­re­al­is­tic. As an anal­ogy, it is well known that the Deep­mind AI al­go­rithm al­phaGO beat the reign­ing GO world cham­pion. What is of­ten over­looked is that the group tasked with de­vel­op­ing the al­phaGO AI al­go­rithm ac­tu­ally had mul­ti­ple GO ex­perts working in the de­sign team.

At Ex­sci­en­tia, we have hu­man drug dis­cov­ery ex­perts working along­side the AI – and the high­est pro­file among these is Andy Bell, a co-in­ven­tor of Viag*ra and other drugs while at Pfizer

Are pharma com­pa­nies ready to adopt these tech­nolo­gies and rev­o­lu­tion­ize health­care?

Most are ex­plor­ing the use of AI now, as ev­i­denced by our own ex­pe­ri­ence and deals. There are many ar­eas in which to ap­ply them so we ex­pect their use to in­crease.

In our own case, de­liv­er­ing con­sis­tently and demon­strat­ing the ef­fi­cien­cies makes other com­pa­nies more likely to ex­plore.

How has AI as­sisted drug dis­cov­ery taken off in Asia?

Our ini­tial fo­cus was solely Ja­pan and Dainip­pon Su­mit­omo was ac­tu­ally one of our first part­ners. Now interest is clearly grow­ing in the broader Asia re­gion, where we see strong po­ten­tial.

Aish­warya Venkatesh aish­warya.venkatesh@mmac­tiv.com

Newspapers in English

Newspapers from India

© PressReader. All rights reserved.