Jacque­line Detwiler on whether all that in­di­vid­u­alised can­cer data can re­ally help pa­tients.

Popular Mechanics (South Africa) - - Contents - BY JACQUE­LINE DETWILER

CASE STUDY 1 The shar­ing project

THREE YEARS AGO, Dr Jinghui Zhang, chair of the depart­ment of com­pu­ta­tional bi­ol­ogy g at St Jude Chil­dren’s Re­search Hos­pi­tal in i Mem­phis, tried to down­load a set of 1 547 sam­ples from a gov­ern­ment data­base so she could run an ex­per­i­ment on it. The down­load took a year and a half. Ev­ery time Zhang’s team thought it had the com­plete set of data, they then per­formed a sta­tis­ti­cal check that was sup­posed to catch any trans­mis­sion er­rors. Seven times, the check failed. Seven times, the team started work­ing only to re­alise that part of the set was miss­ing, and they’d have to start all over. ‘Data down­load­ing is a very painful process,’ Zhang says. ‘It can po­ten­tially drive re­searchers away from sci­ence. Pe­riod.’

Zhang had al­ready started to won­der how much faster can­cer re­search could move if data sets were freely and im­me­di­ately avail­able to the com­pu­ta­tional bi­ol­o­gists who wanted to work on them, so in April she part­nered with Mi­crosoft to re­lease St Jude Cloud, a web-based data pro­ces­sor that of­fers ac­cess – within 48 hours – to the largest pub­licly avail­able set of child­hood can­cer stats in the world, the genomes of more than 5 000 St Jude pa­tients. The shar­ing is so gen­er­ous that it de­fies aca­demic logic. ‘Peo­ple think that there are strings at­tached to this,’ says Zhang. ‘They just don’t be­lieve that we would do such a thing.’

CASE STUDY 2 Ma­chine learn­ing for doc­tors

ST JUDE is also not the only or­gan­i­sa­tion that has seized on shar­ing and col­lab­o­ra­tion as a so­lu­tion to stalled can­cer re­search. In Jan­uary 2018, Mi­crosoft also part­nered with the char­ity Stand Up to Can­cer on an $11 mil­lion (R160 mil­lion) re­search pro­gram called Con­ver­gence 2.0 to solve a prob­lem that is es­sen­tially the re­verse of the one Zhang faced: re­searchers who have data sets or ideas worth study­ing, but lit­tle or no com­puter-pro­gram­ming ex­per­tise. ‘There was hardly a can­cer cen­tre in the world that had the same peo­ple that Face­book, Mi­crosoft, Ama­zon, and Google were get­ting to do deep-learn­ing al­go­rithms. These are two nonover­lap­ping groups of peo­ple,’ says Arnie Levine, pro­fes­sor emer­i­tus at the In­sti­tute for Ad­vanced Study and co-vice chair­per­son of Stand Up to Can­cer’s sci­en­tific ad­vi­sory com­mit­tee. ‘So what we did was we made the mar­riage.’ Con­ver­gence 2.0 funded seven teams of re­searchers and med­i­cal doc­tors, match­ing each with ma­chine-learn­ing ex­perts from places like Mi­crosoft and MIT.

3 CASE STUDY Sharable tu­mour sam­ples

MEAN­WHILE, for clin­i­cians, , a new com­pany called Paige.gai made an agree­ment g with Me­mo­rial Sloan Ket­ter­ing Can­cer Cen­ter in i New York to train ma­chine-learn­ing al­go­rithms on its ab­so­lutely mas­sive col­lec­tion of 25 mil­lion fully digi­tised tu­mour slides. The soft­ware will even­tu­ally be able to help doc­tors all over the coun­try di­ag­nose can­cer from biop­sies the same way a top MSK pathol­o­gist would, only with even more ac­cu­racy, ob­jec­tiv­ity, and in­for­ma­tion about treat­ment op­tions and po­ten­tial sur­vival.

UP­DATE Jacque­line Detwiler also re­ported on the state of can­cer re­search in our Septem­ber 2018 is­sue.

Newspapers in English

Newspapers from South Africa

© PressReader. All rights reserved.