— Ge­netic mugshots

A pri­vate-vs-public de­bate may de­fine the fu­ture of health care

Cosmos - - Contents - PAUL BIEGLER is a philoso­pher, physi­cian and Ad­junct Re­search Fel­low in Bioethics at Monash Univer­sity.

“ORDER. SPIT. DIS­COVER.” That’s a win­ning tagline for the DNA test­ing kit ‘23andme’, a prod­uct so pop­u­lar it has made the top five of Ama­zon’s best-sell­ing prod­ucts.

It’s also an ap­proach ush­er­ing in a new era of per­son­alised medicine. In the­ory, hav­ing your DNA read from a sim­ple saliva sam­ple could be like gaz­ing into a med­i­cal crys­tal ball. It will list the dis­eases you are pre­dis­posed to, help you pre­vent them or guide you to the best treat­ment if you al­ready have them.

For in­stance, let’s say you’re a woman and your test shows you carry the deadly ver­sion of the breast can­cer gene BRCA1. You might have a mas­tec­tomy, as An­gelina Jolie did. Or not. Not ev­ery BRCA1 mu­ta­tion is equally deadly, and it also de­pends on what other genes you carry.

Big data is re­quired to truly re­alise the vi­sion of per­son­alised medicine. We need to con­trib­ute our ge­netic in­for­ma­tion and med­i­cal his­to­ries to data­bases, whose daunt­ing com­plex­ity re­searchers try to de­code – in­creas­ingly with the aid of ma­chine-learn­ing al­go­rithms.

The deal breaker is ge­netic pri­vacy. Those data­bases must be hack-proof if we want to pre­vent un­scrupu­lous in­surance com­pa­nies or em­ploy­ers from lift­ing a per­son’s ge­netic se­crets. The cur­rent stan­dard is to de-iden­tify ge­netic and med­i­cal in­for­ma­tion so there are no linked names or other clues that might be cross­ref­er­enced to trace iden­tity.

In Septem­ber 2017 a star­tling pa­per in the jour­nal PNAS sug­gested ge­netic pri­vacy could no longer be guar­an­teed. The au­thors were from Hu­man Longevity Inc. (HLI), led by its founder Craig Ven­ter, who was one of the sci­en­tists fa­mous for read­ing the hu­man genome in 2001.

HLI claimed it de­ployed a very smart ma­chine-learn­ing al­go­rithm to re­con­struct a per­son’s face, with 80% ac­cu­racy, from a tract of their ge­netic code. Yes, re­ally, a ge­netic mugshot. Throw in fa­cial recog­ni­tion soft­ware and the ubiq­ui­tous Face­book pro­file and HLI might have a dealt a coup de gras to ge­netic pri­vacy.

The re­sult was mind-bend­ing, the sci­en­tific blow­back swift.

Within days, Yaniv Er­lich, a Columbia Univer­sity com­pu­ta­tional bi­ol­o­gist, chal­lenged the pre­dic­tive power of the HLI al­go­rithm. For­mer HLI em­ployee Ja­son Piper, a co-author on the pa­per, went fur­ther by dis­tanc­ing him­self from its con­clu­sions on Twit­ter and ac­cus­ing Ven­ter of craft­ing a re­sult aimed at keep­ing ge­netic data in pri­vate hands.

Piper’s logic: high­light­ing the po­ten­tial to put a face to the genes in a public data­base may lead to the propo­si­tion that guar­an­teed anonymity re­quires keep­ing DNA in the shrouded servers of pri­vate com­pa­nies, such as HLI, which stand to gain plenty.

How be­liev­able is the science? While per­haps not ready for cen­tre stage, it would ap­pear to be hov­er­ing in the wings.

“Can DNA from a scene-of-crime se­men sam­ple give a pic­ture of what a rapist looks like? As of now, prob­a­bly not; as of three years from now, prob­a­bly yes,” says Bob Wil­liamson, hon­orary pro­fes­sor at Mel­bourne’s Mur­doch Chil­dren’s Re­search In­sti­tute.

The ob­vi­ous con­cern is that the rise and rise of ma­chine learn­ing could ring the death knell for ge­netic data shar­ing. How­ever, science is rac­ing to find a so­lu­tion in what is akin to a tech­no­log­i­cal arms race.

US re­searchers re­cently showed a tech­nique called “genome cloak­ing” that can con­ceal most of the ge­netic code, re­veal­ing only a small sub­set of in­ter­est to re­searchers. The true game changer, though, could be block chain en­cryp­tion. Syd­ney’s Gar­van in­sti­tute, for one, has signed tech startup E-nome to se­cure more than 15,000 pa­tient DNA data sets with the tech­nol­ogy.

The stakes are high. In re­cent months, Aus­tralian au­thor­i­ties have pub­lished no fewer than three re­ports about the ben­e­fits of pre­ci­sion medicine. The US Na­tional In­sti­tutes of Health aims to se­quence a mil­lion genomes by 2020.

But the real ac­tion may well lie to our north; China has funded its own pre­ci­sion medicine jug­ger­naut to the tune of US$9 bil­lion. It will team public and pri­vate sec­tors and, on one es­ti­mate, se­quence 100 mil­lion genomes by 2030.

With that vast data­base, China could nail many of pre­ci­sion medicine’s prob­lems – such as the spec­trum of can­cer risk from the BRCA1 gene – at least for its own pop­u­la­tion.

Will China share its in­tel­li­gence with the rest of us? How will the Chi­nese pro­gram nav­i­gate the public-pri­vate de­bate? As it all plays out, the world’s sci­en­tists and ethi­cists will be pay­ing very close at­ten­tion.

Ma­chine learn­ing could ring the death knell for ge­netic data shar­ing.

Newspapers in English

Newspapers from Australia

© PressReader. All rights reserved.