1
00:00:00,480 --> 00:00:03,480
foreign

2
00:00:08,660 --> 00:00:14,880
today we have uh David Lawrence from

3
00:00:12,120 --> 00:00:17,760
South Australia giving us a talk

4
00:00:14,880 --> 00:00:20,640
analyzing and sharing genetic data with

5
00:00:17,760 --> 00:00:22,820
python uh could we give David a warm

6
00:00:20,640 --> 00:00:22,820
welcome

7
00:00:25,280 --> 00:00:29,359
[Applause]

8
00:00:26,820 --> 00:00:29,359
to you David

9
00:00:35,399 --> 00:00:39,420
um good morning

10
00:00:37,860 --> 00:00:42,180
um hi yeah so I have a pretty weird

11
00:00:39,420 --> 00:00:44,340
career um path so I was interested in

12
00:00:42,180 --> 00:00:46,800
computer games at the start and then I

13
00:00:44,340 --> 00:00:48,800
realized I needed more money and not

14
00:00:46,800 --> 00:00:52,200
working insane amount of hours and then

15
00:00:48,800 --> 00:00:55,620
worked in the corporate world doing Java

16
00:00:52,200 --> 00:00:57,120
programming for a while and then went

17
00:00:55,620 --> 00:00:59,460
into bioinformatics and that's actually

18
00:00:57,120 --> 00:01:00,840
been looking at I looked at it today I'm

19
00:00:59,460 --> 00:01:02,460
like wow I've actually been a

20
00:01:00,840 --> 00:01:04,260
bioinformatician longer than a

21
00:01:02,460 --> 00:01:06,360
programmer so that was

22
00:01:04,260 --> 00:01:08,280
pretty strange

23
00:01:06,360 --> 00:01:10,740
um yeah but there's a saying in science

24
00:01:08,280 --> 00:01:12,180
where the best you know as statisticians

25
00:01:10,740 --> 00:01:15,780
they say the best thing about it is you

26
00:01:12,180 --> 00:01:17,460
get to play in other people's backyards

27
00:01:15,780 --> 00:01:19,860
um and I feel the same about programmers

28
00:01:17,460 --> 00:01:21,180
so you know um if you want to if you a

29
00:01:19,860 --> 00:01:23,340
programmer you can

30
00:01:21,180 --> 00:01:24,360
um you know you can end up working on a

31
00:01:23,340 --> 00:01:27,060
national

32
00:01:24,360 --> 00:01:28,500
um electricity Grid or genetics you know

33
00:01:27,060 --> 00:01:30,060
so there's a lot of cool things we can

34
00:01:28,500 --> 00:01:32,159
do

35
00:01:30,060 --> 00:01:33,360
um so the place I get to play is in

36
00:01:32,159 --> 00:01:35,220
biology

37
00:01:33,360 --> 00:01:37,439
um which I I'm biased but I think it's

38
00:01:35,220 --> 00:01:39,119
one of the most interesting things uh

39
00:01:37,439 --> 00:01:41,220
there is

40
00:01:39,119 --> 00:01:43,439
um so everyone you know because of the

41
00:01:41,220 --> 00:01:48,600
vaccines most people know about mRNA now

42
00:01:43,439 --> 00:01:51,000
so DNA makes RNA which makes protein and

43
00:01:48,600 --> 00:01:53,520
they've actually you know it was only in

44
00:01:51,000 --> 00:01:55,200
my grandparents era so maybe their sixth

45
00:01:53,520 --> 00:01:57,479
or my parents actually in the 60s they

46
00:01:55,200 --> 00:01:59,579
discovered how the DNA gets turned into

47
00:01:57,479 --> 00:02:00,860
protein so there's actually a code it's

48
00:01:59,579 --> 00:02:04,680
called codons

49
00:02:00,860 --> 00:02:07,380
and the the bases get changed um the U

50
00:02:04,680 --> 00:02:11,180
is the RNA version of a t but yeah

51
00:02:07,380 --> 00:02:14,400
basically this is how it's translated

52
00:02:11,180 --> 00:02:17,400
and basically the DNA gets translated

53
00:02:14,400 --> 00:02:19,319
three at a time into proteins and then

54
00:02:17,400 --> 00:02:20,760
proteins they fold up that's a

55
00:02:19,319 --> 00:02:23,580
computational problem that people are

56
00:02:20,760 --> 00:02:25,500
working on and it folds not just a

57
00:02:23,580 --> 00:02:26,459
single protein it folds into multiple

58
00:02:25,500 --> 00:02:29,000
ones

59
00:02:26,459 --> 00:02:31,680
then into complexes and then it gets

60
00:02:29,000 --> 00:02:33,900
ridiculously complicated and hard so

61
00:02:31,680 --> 00:02:36,540
there's a lot of computational stuff to

62
00:02:33,900 --> 00:02:38,760
do in biology

63
00:02:36,540 --> 00:02:40,020
um so yeah and bioinformatics when you

64
00:02:38,760 --> 00:02:43,620
think about it it the business

65
00:02:40,020 --> 00:02:46,800
requirement is basically to discover all

66
00:02:43,620 --> 00:02:48,420
about this stuff about life and we know

67
00:02:46,800 --> 00:02:50,160
a lot but there's

68
00:02:48,420 --> 00:02:51,900
an astonishing amount that we don't know

69
00:02:50,160 --> 00:02:56,400
there's definitely lifetimes worth of

70
00:02:51,900 --> 00:02:58,400
work to to find it all

71
00:02:56,400 --> 00:03:00,599
um yeah so there's a big demand for

72
00:02:58,400 --> 00:03:03,180
programmers and the more python

73
00:03:00,599 --> 00:03:05,040
programmers in bioinformatics the better

74
00:03:03,180 --> 00:03:07,200
I think the better it needs are some

75
00:03:05,040 --> 00:03:08,340
more a lot of software skills there's a

76
00:03:07,200 --> 00:03:10,739
lot of people are sort of self-taught

77
00:03:08,340 --> 00:03:13,500
biologists and um yeah if you can come

78
00:03:10,739 --> 00:03:16,260
in with um the software skills you'd be

79
00:03:13,500 --> 00:03:17,099
very welcome so yeah check out the jobs

80
00:03:16,260 --> 00:03:19,440
um

81
00:03:17,099 --> 00:03:20,700
yeah um and if you're interested my

82
00:03:19,440 --> 00:03:23,640
recommendation for getting into

83
00:03:20,700 --> 00:03:25,140
bioinformatics is this um online web app

84
00:03:23,640 --> 00:03:27,959
it's basically leak code for

85
00:03:25,140 --> 00:03:29,760
bioinformatics project Rosalind

86
00:03:27,959 --> 00:03:33,180
um and yeah basically it starts off

87
00:03:29,760 --> 00:03:34,560
really simple counting nucleotides so

88
00:03:33,180 --> 00:03:36,599
you basically just have to they randomly

89
00:03:34,560 --> 00:03:38,420
generate an output you run your Python

90
00:03:36,599 --> 00:03:40,560
program or whatever program against it

91
00:03:38,420 --> 00:03:42,420
paste the answer in if you got it right

92
00:03:40,560 --> 00:03:46,500
you move on to the next one starts off

93
00:03:42,420 --> 00:03:48,900
easy gets really hard so lots of fun so

94
00:03:46,500 --> 00:03:52,260
I'm going to talk about my particular

95
00:03:48,900 --> 00:03:54,120
job so this is I work in pathology so

96
00:03:52,260 --> 00:03:57,000
here's an example of something that that

97
00:03:54,120 --> 00:03:59,519
happens to me so basically in well me

98
00:03:57,000 --> 00:04:00,780
and a massive team so in the Women's and

99
00:03:59,519 --> 00:04:01,980
Children's Hospital the other side of

100
00:04:00,780 --> 00:04:03,720
the river

101
00:04:01,980 --> 00:04:05,519
um generally people you know a baby

102
00:04:03,720 --> 00:04:07,319
might be born with heart problems and

103
00:04:05,519 --> 00:04:10,080
the question is is it due to infection

104
00:04:07,319 --> 00:04:12,299
is it due to a genetic predisposition is

105
00:04:10,080 --> 00:04:15,360
it due to birth complications all kinds

106
00:04:12,299 --> 00:04:18,959
of troubles and knowing the reason

107
00:04:15,360 --> 00:04:22,079
um helps the clinician decide what to do

108
00:04:18,959 --> 00:04:24,060
so how pathology works is basically the

109
00:04:22,079 --> 00:04:26,520
clinician sends an order to the

110
00:04:24,060 --> 00:04:29,100
Pathology company essay pathology is the

111
00:04:26,520 --> 00:04:32,220
public service provider for pathology in

112
00:04:29,100 --> 00:04:34,500
South Australia we take blood we take

113
00:04:32,220 --> 00:04:37,020
information and we send back a report

114
00:04:34,500 --> 00:04:39,120
that informs the clinician's decisions

115
00:04:37,020 --> 00:04:41,580
on what to do

116
00:04:39,120 --> 00:04:43,560
so there's that you know that's how it

117
00:04:41,580 --> 00:04:45,780
worked for things like um you know blood

118
00:04:43,560 --> 00:04:47,699
concentrations of some you know calcium

119
00:04:45,780 --> 00:04:50,160
or something it's a bit more complicated

120
00:04:47,699 --> 00:04:52,440
with DNA sequencing so we still take

121
00:04:50,160 --> 00:04:54,840
blood we extract the blood

122
00:04:52,440 --> 00:04:57,300
um put it in a DNA sequencer process it

123
00:04:54,840 --> 00:05:00,180
on HPC machines work out the difference

124
00:04:57,300 --> 00:05:02,360
between like a reference human genome

125
00:05:00,180 --> 00:05:05,520
and and this patient's Human Genome

126
00:05:02,360 --> 00:05:07,800
analyze it and send back the report so

127
00:05:05,520 --> 00:05:10,440
this is at foam Road the other side of

128
00:05:07,800 --> 00:05:14,100
um North Terrace this is my colleague

129
00:05:10,440 --> 00:05:17,639
and our expensive new sequencer the way

130
00:05:14,100 --> 00:05:21,419
it works is basically the DNA gets put

131
00:05:17,639 --> 00:05:23,639
on this uh sort of glass tile and they

132
00:05:21,419 --> 00:05:27,259
sort of wash the whole thing in bases

133
00:05:23,639 --> 00:05:29,460
and when the next base at a time gets um

134
00:05:27,259 --> 00:05:30,720
Incorporated and it flashes a certain

135
00:05:29,460 --> 00:05:32,340
color and there's a camera in the

136
00:05:30,720 --> 00:05:35,340
machine and it just like

137
00:05:32,340 --> 00:05:37,440
um basically sees that you know there's

138
00:05:35,340 --> 00:05:39,840
a million flashes of this color and that

139
00:05:37,440 --> 00:05:43,380
color and it works out what it is and

140
00:05:39,840 --> 00:05:45,660
eventually it turns it into DNA

141
00:05:43,380 --> 00:05:48,240
sequences so this is the raw output of a

142
00:05:45,660 --> 00:05:51,240
sequencer the new one we have is 16 to

143
00:05:48,240 --> 00:05:54,960
20 billion reads in 44 hours and these

144
00:05:51,240 --> 00:05:56,940
are usually like 250 bases long the

145
00:05:54,960 --> 00:06:00,000
other things are like quality score and

146
00:05:56,940 --> 00:06:02,460
how confident the base calling is

147
00:06:00,000 --> 00:06:05,039
so what do we do how do we you know 200

148
00:06:02,460 --> 00:06:08,220
bases where you know what does that mean

149
00:06:05,039 --> 00:06:10,680
um see the 200 bases we work out where

150
00:06:08,220 --> 00:06:12,539
in the three billion human bases that

151
00:06:10,680 --> 00:06:14,340
came from so there's like a string

152
00:06:12,539 --> 00:06:15,960
matching with a bit of fuzziness in

153
00:06:14,340 --> 00:06:18,900
there where we map to the human

154
00:06:15,960 --> 00:06:21,660
reference genome so

155
00:06:18,900 --> 00:06:25,740
um and here's an example of a mapping

156
00:06:21,660 --> 00:06:28,199
um the uh these all there's no cut no

157
00:06:25,740 --> 00:06:31,220
letters in the map reads that means it's

158
00:06:28,199 --> 00:06:34,860
a perfect match but the G in the middle

159
00:06:31,220 --> 00:06:37,259
that's a mismatch so the patient differs

160
00:06:34,860 --> 00:06:39,960
from the reference genome the reference

161
00:06:37,259 --> 00:06:42,720
genome has an a the patient has a g half

162
00:06:39,960 --> 00:06:44,840
of the time so you have two copies of

163
00:06:42,720 --> 00:06:47,400
chromosomes one from mum one from Dad

164
00:06:44,840 --> 00:06:49,440
and the on one of them we don't know

165
00:06:47,400 --> 00:06:51,960
which either the mum's copy or the dad's

166
00:06:49,440 --> 00:06:55,139
copy um there's a g instead of an A so

167
00:06:51,960 --> 00:06:57,240
this is called a heterozygous mutation

168
00:06:55,139 --> 00:06:59,940
so where do these human genomes come

169
00:06:57,240 --> 00:07:02,280
from uh it's basically was it extremely

170
00:06:59,940 --> 00:07:04,020
expensive project in the early 2000s and

171
00:07:02,280 --> 00:07:05,220
yeah you can basically download this off

172
00:07:04,020 --> 00:07:07,199
the internet

173
00:07:05,220 --> 00:07:08,340
um through you know it cost insane

174
00:07:07,199 --> 00:07:10,680
amount of money to produce you can

175
00:07:08,340 --> 00:07:12,419
download a text file I use it we use it

176
00:07:10,680 --> 00:07:14,699
in genomics for everything uh it's

177
00:07:12,419 --> 00:07:17,460
extremely useful thanks for paying for

178
00:07:14,699 --> 00:07:19,500
science back you know 23 years ago I

179
00:07:17,460 --> 00:07:21,120
will use it every day

180
00:07:19,500 --> 00:07:22,979
um and Yeah the secret you know it was

181
00:07:21,120 --> 00:07:25,979
insanely expensive to do sequencing in

182
00:07:22,979 --> 00:07:27,780
the past but the it's actually much uh

183
00:07:25,979 --> 00:07:30,120
falling faster than Moore's Law so if

184
00:07:27,780 --> 00:07:32,940
you think computers are getting uh

185
00:07:30,120 --> 00:07:36,539
faster over time well uh genomics is uh

186
00:07:32,940 --> 00:07:38,819
going even crazier and when things get

187
00:07:36,539 --> 00:07:41,819
cheaper people want more of it this is

188
00:07:38,819 --> 00:07:43,440
our demand for the stuff running through

189
00:07:41,819 --> 00:07:45,900
our software

190
00:07:43,440 --> 00:07:48,060
um that starts around about 2013 and

191
00:07:45,900 --> 00:07:51,180
it's getting nuts right so

192
00:07:48,060 --> 00:07:54,360
good times so how's python used in

193
00:07:51,180 --> 00:07:56,539
bioinformatics well generally uh the

194
00:07:54,360 --> 00:08:00,419
tools for doing the uh the hardcore

195
00:07:56,539 --> 00:08:03,060
computational stuff are written in C for

196
00:08:00,419 --> 00:08:06,660
Speed or sometimes rust now but pythons

197
00:08:03,060 --> 00:08:09,000
very commonly used to sort of uh join

198
00:08:06,660 --> 00:08:14,220
the multiple tools together so you might

199
00:08:09,000 --> 00:08:15,960
Define a workflow and coordinate the

200
00:08:14,220 --> 00:08:18,599
running of all these different jobs in

201
00:08:15,960 --> 00:08:19,860
Python so here's an example I'm just

202
00:08:18,599 --> 00:08:21,240
going to skip over this but there's a

203
00:08:19,860 --> 00:08:24,120
program called snake make which is

204
00:08:21,240 --> 00:08:26,879
basically like make but it allows python

205
00:08:24,120 --> 00:08:28,620
so it's pretty cool using bioinformatics

206
00:08:26,879 --> 00:08:32,580
maybe it's cool useful for you guys as

207
00:08:28,620 --> 00:08:35,599
well for dependency stuff

208
00:08:32,580 --> 00:08:38,219
um so for research typically

209
00:08:35,599 --> 00:08:39,839
researchers work in small teams and the

210
00:08:38,219 --> 00:08:42,360
output is basically a paper so someone

211
00:08:39,839 --> 00:08:44,159
has a biological question and you work

212
00:08:42,360 --> 00:08:46,920
with the biologist for six months to

213
00:08:44,159 --> 00:08:48,959
three years and a lot of the time it's

214
00:08:46,920 --> 00:08:51,899
stuff like we have this data we have a

215
00:08:48,959 --> 00:08:53,880
biological question what can we do and

216
00:08:51,899 --> 00:08:56,519
you often try a whole bunch of stuff you

217
00:08:53,880 --> 00:08:58,860
write one-off programs you investigate

218
00:08:56,519 --> 00:09:02,220
it and python is amazing for this

219
00:08:58,860 --> 00:09:03,899
because you know I can write python 10

220
00:09:02,220 --> 00:09:05,880
times faster than C or something I don't

221
00:09:03,899 --> 00:09:07,500
know what it is but it's like it's so

222
00:09:05,880 --> 00:09:10,019
much easier and and the thing is we have

223
00:09:07,500 --> 00:09:12,060
HPC we can run we've got hundreds of

224
00:09:10,019 --> 00:09:13,500
cores we can run things in parallel it

225
00:09:12,060 --> 00:09:14,940
doesn't matter how fast it is often we

226
00:09:13,500 --> 00:09:17,580
only run it once we're just

227
00:09:14,940 --> 00:09:19,320
experimenting and if we can experiment

228
00:09:17,580 --> 00:09:21,720
twice in a day rather than once every

229
00:09:19,320 --> 00:09:25,019
few days we converge on the answer

230
00:09:21,720 --> 00:09:26,880
faster and that's great

231
00:09:25,019 --> 00:09:29,880
um so yeah to go back to varying tools

232
00:09:26,880 --> 00:09:31,200
when so the original sequence is at the

233
00:09:29,880 --> 00:09:33,380
top and the bottom one is at the bottom

234
00:09:31,200 --> 00:09:36,060
that's called in point mutation

235
00:09:33,380 --> 00:09:39,000
we have a file format for this it's

236
00:09:36,060 --> 00:09:42,360
called the VCF format and basically

237
00:09:39,000 --> 00:09:44,880
there's a chromosome location a

238
00:09:42,360 --> 00:09:47,040
chromosome name a position and then the

239
00:09:44,880 --> 00:09:50,459
original sequence and then the new

240
00:09:47,040 --> 00:09:52,440
sequence and a bunch of other stuff so

241
00:09:50,459 --> 00:09:54,959
um I've written some software that some

242
00:09:52,440 --> 00:09:58,680
a user essay path to analyze this kind

243
00:09:54,959 --> 00:10:01,440
of data and it's called variant grid

244
00:09:58,680 --> 00:10:03,060
um so what we do is

245
00:10:01,440 --> 00:10:04,920
um we start with the sequencing which

246
00:10:03,060 --> 00:10:06,779
has a huge amount of data we run then we

247
00:10:04,920 --> 00:10:10,680
do the get the diffs and we do all this

248
00:10:06,779 --> 00:10:14,100
on an HPC and then from this

249
00:10:10,680 --> 00:10:16,320
um the VCF for the variant calls then it

250
00:10:14,100 --> 00:10:20,399
gets ingested into our program and I

251
00:10:16,320 --> 00:10:23,160
take over from that point there

252
00:10:20,399 --> 00:10:25,380
um so it's basically a Django app

253
00:10:23,160 --> 00:10:28,100
um it's running on g-unicorn with

254
00:10:25,380 --> 00:10:30,420
postgres being the central database

255
00:10:28,100 --> 00:10:32,519
and whenever we have anything you know

256
00:10:30,420 --> 00:10:34,980
two more than a couple of seconds I farm

257
00:10:32,519 --> 00:10:38,459
it off to a um to a queue and salary

258
00:10:34,980 --> 00:10:39,899
workers do the work most of the time the

259
00:10:38,459 --> 00:10:42,480
the longest running jobs might be say

260
00:10:39,899 --> 00:10:44,279
half an hour which is for annotations

261
00:10:42,480 --> 00:10:46,440
they run things like computational

262
00:10:44,279 --> 00:10:49,920
prediction so for instance you might

263
00:10:46,440 --> 00:10:53,399
have a sequence and it'll do things like

264
00:10:49,920 --> 00:10:56,579
um look at the charges of the the new

265
00:10:53,399 --> 00:10:58,980
modified protein and work out how much

266
00:10:56,579 --> 00:11:00,959
of a sort of how wonky the protein will

267
00:10:58,980 --> 00:11:02,399
go and classify it as pathogenic or

268
00:11:00,959 --> 00:11:04,260
something and so there's all these

269
00:11:02,399 --> 00:11:07,260
different tools and public databases

270
00:11:04,260 --> 00:11:08,700
that we use to work out what a variant

271
00:11:07,260 --> 00:11:11,579
does

272
00:11:08,700 --> 00:11:15,060
so yeah here's the VCF again

273
00:11:11,579 --> 00:11:17,040
um and uh so the way I represent it is

274
00:11:15,060 --> 00:11:18,600
in Django

275
00:11:17,040 --> 00:11:20,399
um yeah because this isn't in the Django

276
00:11:18,600 --> 00:11:21,300
stream not sure if everyone knows Django

277
00:11:20,399 --> 00:11:23,399
but

278
00:11:21,300 --> 00:11:25,019
um basically you can declare classes

279
00:11:23,399 --> 00:11:27,120
that represent

280
00:11:25,019 --> 00:11:30,120
um database tables

281
00:11:27,120 --> 00:11:32,839
um so here's my Chrome contig is just

282
00:11:30,120 --> 00:11:35,160
another biology name for a chromosome

283
00:11:32,839 --> 00:11:37,200
position reference and there might be

284
00:11:35,160 --> 00:11:38,220
multiple variants for a location like

285
00:11:37,200 --> 00:11:40,079
you know

286
00:11:38,220 --> 00:11:42,260
people might not have a t you might have

287
00:11:40,079 --> 00:11:44,940
a c someone else might have a g

288
00:11:42,260 --> 00:11:47,240
and so you can have multiple ones there

289
00:11:44,940 --> 00:11:49,800
and you can write your python classes

290
00:11:47,240 --> 00:11:52,140
and it sort of handles the SQL for you

291
00:11:49,800 --> 00:11:56,160
so instead like on the bottom right here

292
00:11:52,140 --> 00:11:58,740
is the database and on the top left is

293
00:11:56,160 --> 00:12:00,480
how you write the Django code and it

294
00:11:58,740 --> 00:12:02,880
means that instead of writing all the

295
00:12:00,480 --> 00:12:04,320
you know technical computery stuff

296
00:12:02,880 --> 00:12:06,839
you're just sort of working with

297
00:12:04,320 --> 00:12:10,140
variants and samples and you you end up

298
00:12:06,839 --> 00:12:12,600
sort of working in your domain which is

299
00:12:10,140 --> 00:12:13,940
really useful instead of computer you

300
00:12:12,600 --> 00:12:15,480
know code

301
00:12:13,940 --> 00:12:17,459
but

302
00:12:15,480 --> 00:12:18,060
um yeah so there's a huge amount of

303
00:12:17,459 --> 00:12:20,399
um

304
00:12:18,060 --> 00:12:23,100
biological data and we get basically

305
00:12:20,399 --> 00:12:24,720
everything that we can find we'll we'll

306
00:12:23,100 --> 00:12:27,480
run it and that you know there's

307
00:12:24,720 --> 00:12:29,700
hundreds we have hundreds of columns of

308
00:12:27,480 --> 00:12:33,420
different tools and public databases

309
00:12:29,700 --> 00:12:36,240
that tell us what a variant does

310
00:12:33,420 --> 00:12:37,740
um so yeah there ends up being uh uh you

311
00:12:36,240 --> 00:12:39,839
know hundreds of thousands to millions

312
00:12:37,740 --> 00:12:42,300
of variants um and almost all you know

313
00:12:39,839 --> 00:12:44,579
they're things like be a millimeter

314
00:12:42,300 --> 00:12:46,740
taller or shorter or have um a slightly

315
00:12:44,579 --> 00:12:48,000
different color hair or something so you

316
00:12:46,740 --> 00:12:51,240
don't care about that when you're

317
00:12:48,000 --> 00:12:53,399
searching for the um genetic cause of

318
00:12:51,240 --> 00:12:57,300
disease so most of the work is throwing

319
00:12:53,399 --> 00:12:59,100
out all of that stuff and to fight to

320
00:12:57,300 --> 00:13:00,899
search through this it's not easy you

321
00:12:59,100 --> 00:13:02,339
have to know um it's like a mini

322
00:13:00,899 --> 00:13:04,860
research project you have to know about

323
00:13:02,339 --> 00:13:07,860
the disease how common it is the family

324
00:13:04,860 --> 00:13:10,740
history you end up going and doing uh

325
00:13:07,860 --> 00:13:12,720
looking through the literature and so

326
00:13:10,740 --> 00:13:14,399
these met these team of people called

327
00:13:12,720 --> 00:13:17,279
medical scientists who basically do this

328
00:13:14,399 --> 00:13:19,320
investigation and the goal is basically

329
00:13:17,279 --> 00:13:22,440
of this program is to allow those people

330
00:13:19,320 --> 00:13:23,820
to run their own filters on huge amounts

331
00:13:22,440 --> 00:13:26,220
of data sets because they've got

332
00:13:23,820 --> 00:13:27,360
excellent medical knowledge but not so

333
00:13:26,220 --> 00:13:29,880
you know they're not going to be able to

334
00:13:27,360 --> 00:13:33,480
run Jupiter notebooks or something

335
00:13:29,880 --> 00:13:35,220
so a naive way to do filtering would be

336
00:13:33,480 --> 00:13:37,320
something like this where you basically

337
00:13:35,220 --> 00:13:39,240
build up filters but the trouble is

338
00:13:37,320 --> 00:13:41,639
we're offering off and running something

339
00:13:39,240 --> 00:13:44,160
like 50 to 100 filters and this gets

340
00:13:41,639 --> 00:13:47,279
unwieldy very fast

341
00:13:44,160 --> 00:13:49,139
so basically I have a mate who works um

342
00:13:47,279 --> 00:13:53,579
in video compositing and he showed me

343
00:13:49,139 --> 00:13:55,800
one day a directed a cyclic graph often

344
00:13:53,579 --> 00:13:57,360
these are sort of tilted sideways and

345
00:13:55,800 --> 00:13:58,620
basically the output of one node goes

346
00:13:57,360 --> 00:14:02,220
into the other one some kind of

347
00:13:58,620 --> 00:14:04,680
filtering or thing happens and that's

348
00:14:02,220 --> 00:14:08,000
how it works and that allows people with

349
00:14:04,680 --> 00:14:10,920
domain knowledge to apply computational

350
00:14:08,000 --> 00:14:14,180
work which is exactly what we want to do

351
00:14:10,920 --> 00:14:16,680
so ripped it off and do it for genomics

352
00:14:14,180 --> 00:14:19,220
so the way it works under the hood is

353
00:14:16,680 --> 00:14:21,779
each node returns a Django Q object

354
00:14:19,220 --> 00:14:24,839
which is basically Django's way of

355
00:14:21,779 --> 00:14:27,120
filtering a query set or an SQL query so

356
00:14:24,839 --> 00:14:31,820
you can take these Q objects do logical

357
00:14:27,120 --> 00:14:35,639
operations on them and make a SQL query

358
00:14:31,820 --> 00:14:38,339
so here's an example so I have

359
00:14:35,639 --> 00:14:39,779
um here um there's populate like I've

360
00:14:38,339 --> 00:14:41,940
got a whole bunch of queue objects one

361
00:14:39,779 --> 00:14:44,820
of them is population frequency greater

362
00:14:41,940 --> 00:14:46,800
than or equal to 0.01 one of them is

363
00:14:44,820 --> 00:14:49,440
Gene symbol in this Gene list and the

364
00:14:46,800 --> 00:14:52,440
other one is damage greater than 0.0.5

365
00:14:49,440 --> 00:14:53,940
and then you can run

366
00:14:52,440 --> 00:14:58,440
um you know you could write this by

367
00:14:53,940 --> 00:15:02,459
saying you know q and Q 2 and Q3 but you

368
00:14:58,440 --> 00:15:05,220
can also just reduce it down applying

369
00:15:02,459 --> 00:15:07,380
and like that and you can if you want to

370
00:15:05,220 --> 00:15:09,360
have basically you know apply this

371
00:15:07,380 --> 00:15:11,339
filter or this filter or this filter and

372
00:15:09,360 --> 00:15:14,519
this filter you can build all kind of

373
00:15:11,339 --> 00:15:16,079
logical operations in this way

374
00:15:14,519 --> 00:15:19,620
and this is what the GUI looks like

375
00:15:16,079 --> 00:15:22,139
basically on the left is the nodes the

376
00:15:19,620 --> 00:15:23,639
filters and you allow and you click on

377
00:15:22,139 --> 00:15:26,519
one so we've clicked on the red one in

378
00:15:23,639 --> 00:15:29,040
the middle and that loads the population

379
00:15:26,519 --> 00:15:30,839
frequency and you can choose what

380
00:15:29,040 --> 00:15:33,120
databases you want to use in your

381
00:15:30,839 --> 00:15:36,320
population frequency and below that is

382
00:15:33,120 --> 00:15:39,899
the grid of the variance like the 31

383
00:15:36,320 --> 00:15:42,000
that appear in that node and it allows

384
00:15:39,899 --> 00:15:44,040
the sort of the medical scientist to

385
00:15:42,000 --> 00:15:46,199
sort of jump around and like apply more

386
00:15:44,040 --> 00:15:47,820
and more or less stringent filters if

387
00:15:46,199 --> 00:15:49,800
something comes through or not they can

388
00:15:47,820 --> 00:15:52,380
adjust the population frequency if it's

389
00:15:49,800 --> 00:15:54,540
not quite known how common is it disease

390
00:15:52,380 --> 00:15:56,040
is things like that so it allows them to

391
00:15:54,540 --> 00:15:58,740
do their own filtering

392
00:15:56,040 --> 00:16:00,779
uh and yeah we try and allow the GUI to

393
00:15:58,740 --> 00:16:02,339
work nicely for

394
00:16:00,779 --> 00:16:03,240
um the particular filters they want to

395
00:16:02,339 --> 00:16:05,880
run

396
00:16:03,240 --> 00:16:08,279
and if you click on a variant behind

397
00:16:05,880 --> 00:16:10,980
this loads a page per variant and we

398
00:16:08,279 --> 00:16:13,339
have all that annotation details um that

399
00:16:10,980 --> 00:16:13,339
you can view

400
00:16:13,440 --> 00:16:17,040
um yeah because the data size we have

401
00:16:15,120 --> 00:16:19,680
something like 30 000 samples each with

402
00:16:17,040 --> 00:16:23,880
a million rows we store everything in

403
00:16:19,680 --> 00:16:25,860
per patient part uh postgres partitions

404
00:16:23,880 --> 00:16:28,440
and we also filter those we might have

405
00:16:25,860 --> 00:16:30,839
multiple partitions per sample so 95 of

406
00:16:28,440 --> 00:16:33,300
a patient variants are common but a lot

407
00:16:30,839 --> 00:16:34,740
of the time if it's a rare disease you

408
00:16:33,300 --> 00:16:36,899
know it happens one in ten thousand

409
00:16:34,740 --> 00:16:39,660
people you just want to throw away those

410
00:16:36,899 --> 00:16:42,060
really common variants so by putting

411
00:16:39,660 --> 00:16:43,980
them in separate partitions we can

412
00:16:42,060 --> 00:16:45,779
choose to jump to just the partition

413
00:16:43,980 --> 00:16:49,740
we're interested in which makes it

414
00:16:45,779 --> 00:16:52,259
really fast so here's an example where

415
00:16:49,740 --> 00:16:54,660
at the start We have basically

416
00:16:52,259 --> 00:16:56,699
collections is the name of the

417
00:16:54,660 --> 00:16:59,339
partitions that we're after so we always

418
00:16:56,699 --> 00:17:01,339
apply the rare one but then if someone

419
00:16:59,339 --> 00:17:04,380
asks for common variance which means

420
00:17:01,339 --> 00:17:05,780
they haven't applied the you know

421
00:17:04,380 --> 00:17:08,160
population filter

422
00:17:05,780 --> 00:17:10,679
then we also include the common

423
00:17:08,160 --> 00:17:14,040
collection which is like uh 20 times as

424
00:17:10,679 --> 00:17:16,439
big and then we do a filtered relation

425
00:17:14,040 --> 00:17:18,540
which is Django's way of joining to a

426
00:17:16,439 --> 00:17:21,799
table with a constraint and because of

427
00:17:18,540 --> 00:17:24,839
that constraint like we say just this

428
00:17:21,799 --> 00:17:27,780
parent table just this one and then the

429
00:17:24,839 --> 00:17:29,340
postgres knows okay well because of that

430
00:17:27,780 --> 00:17:31,200
constraint I only have to look in this

431
00:17:29,340 --> 00:17:33,240
partition and that makes the query

432
00:17:31,200 --> 00:17:35,100
really fast because yeah if we if we

433
00:17:33,240 --> 00:17:37,140
didn't jump to the partitions

434
00:17:35,100 --> 00:17:39,120
we're talking about billions and

435
00:17:37,140 --> 00:17:41,520
billions of rows and that's not going to

436
00:17:39,120 --> 00:17:43,320
perform in real time

437
00:17:41,520 --> 00:17:45,240
so once we've found these variants we

438
00:17:43,320 --> 00:17:48,720
have classification and this is the

439
00:17:45,240 --> 00:17:50,220
classification uh part of the project so

440
00:17:48,720 --> 00:17:53,160
basically we Auto populated but the

441
00:17:50,220 --> 00:17:54,960
medical scientists can also tweak it go

442
00:17:53,160 --> 00:17:56,580
grab information from various other

443
00:17:54,960 --> 00:17:59,280
sources some things you can't automate

444
00:17:56,580 --> 00:18:01,020
like interpreting literature uh well not

445
00:17:59,280 --> 00:18:03,299
yet

446
00:18:01,020 --> 00:18:05,640
um and yeah so ultimately what they do

447
00:18:03,299 --> 00:18:07,860
is they follow a framework called the

448
00:18:05,640 --> 00:18:10,020
acmg guidelines and they finally

449
00:18:07,860 --> 00:18:12,480
classify a variant in one of five ways

450
00:18:10,020 --> 00:18:15,840
it's benign likely benign

451
00:18:12,480 --> 00:18:18,240
um uncertain and finally likely and then

452
00:18:15,840 --> 00:18:21,000
pathogenic that means definitely disease

453
00:18:18,240 --> 00:18:23,820
causing and then we send that back to

454
00:18:21,000 --> 00:18:26,400
the clinicians and they can look up

455
00:18:23,820 --> 00:18:30,120
drugs and stuff to use

456
00:18:26,400 --> 00:18:33,419
so one of the troubles with this is that

457
00:18:30,120 --> 00:18:35,640
um the state uh the health in Australia

458
00:18:33,419 --> 00:18:37,140
is basically run by States and so

459
00:18:35,640 --> 00:18:39,480
everyone sort of works by themselves and

460
00:18:37,140 --> 00:18:41,039
don't talk to each other so if someone

461
00:18:39,480 --> 00:18:43,020
in Victoria came in and there was a

462
00:18:41,039 --> 00:18:45,419
variant that was pathogenic and someone

463
00:18:43,020 --> 00:18:47,880
in South Australia and they had the same

464
00:18:45,419 --> 00:18:49,740
one but they said it was unknown

465
00:18:47,880 --> 00:18:52,140
um then they would never we would never

466
00:18:49,740 --> 00:18:54,480
know and someone one of them got a bad

467
00:18:52,140 --> 00:18:56,640
diagnosis

468
00:18:54,480 --> 00:18:58,799
um so ideally we'd like to share but

469
00:18:56,640 --> 00:19:01,320
it's actually really difficult because

470
00:18:58,799 --> 00:19:03,480
um it's patient data and there's all

471
00:19:01,320 --> 00:19:05,640
kinds of ethical and security issues to

472
00:19:03,480 --> 00:19:07,559
do to deal with that

473
00:19:05,640 --> 00:19:09,840
um so there's a group called Australian

474
00:19:07,559 --> 00:19:12,360
genomics and they sort of sit as a

475
00:19:09,840 --> 00:19:14,820
national group um that sort of works

476
00:19:12,360 --> 00:19:18,000
above the states to try and help the

477
00:19:14,820 --> 00:19:21,059
individual states work better especially

478
00:19:18,000 --> 00:19:24,000
better together so they came up with the

479
00:19:21,059 --> 00:19:26,000
the plan of coming of sharians which is

480
00:19:24,000 --> 00:19:28,860
basically they have a central system

481
00:19:26,000 --> 00:19:31,140
where you take the state's individual a

482
00:19:28,860 --> 00:19:33,320
class classifications and send them up

483
00:19:31,140 --> 00:19:36,179
to a central server with all of the

484
00:19:33,320 --> 00:19:39,240
anonymity and all kinds of stuff that

485
00:19:36,179 --> 00:19:40,980
you have to do with and if there's a

486
00:19:39,240 --> 00:19:42,720
discrepancy so two different states have

487
00:19:40,980 --> 00:19:45,240
different results then it basically

488
00:19:42,720 --> 00:19:46,380
sends an email and says look sort this

489
00:19:45,240 --> 00:19:48,299
out

490
00:19:46,380 --> 00:19:50,760
and then there's a little like workflow

491
00:19:48,299 --> 00:19:52,260
that we send them down

492
00:19:50,760 --> 00:19:54,179
um so yeah this um Project's been

493
00:19:52,260 --> 00:19:55,919
running since 2019 and so far we've

494
00:19:54,179 --> 00:19:59,039
Linked UP basically every public

495
00:19:55,919 --> 00:20:01,620
pathology provider in Australia and

496
00:19:59,039 --> 00:20:03,960
um also just recently New Zealand we're

497
00:20:01,620 --> 00:20:07,980
now starting to go after

498
00:20:03,960 --> 00:20:10,460
um the uh private Labs pathology labs

499
00:20:07,980 --> 00:20:13,260
and we're also starting to go into

500
00:20:10,460 --> 00:20:15,720
collecting the cancer ones which are a

501
00:20:13,260 --> 00:20:17,520
little bit different and require a bit

502
00:20:15,720 --> 00:20:19,679
so the ones we've traditionally captured

503
00:20:17,520 --> 00:20:21,299
are the inherited ones and we're

504
00:20:19,679 --> 00:20:22,919
starting to capture the cancer ones

505
00:20:21,299 --> 00:20:24,780
which are have a bit of different

506
00:20:22,919 --> 00:20:27,480
information

507
00:20:24,780 --> 00:20:28,919
so the way it works is basically so

508
00:20:27,480 --> 00:20:32,640
essay pathology you know they use

509
00:20:28,919 --> 00:20:35,460
variant grid and so we can do whatever

510
00:20:32,640 --> 00:20:37,620
we like there but in other states they

511
00:20:35,460 --> 00:20:39,720
use different software and so we sort of

512
00:20:37,620 --> 00:20:41,580
have to deal like right integration

513
00:20:39,720 --> 00:20:45,179
software with them and talk to their

514
00:20:41,580 --> 00:20:47,700
system and then translate the data from

515
00:20:45,179 --> 00:20:52,080
their format into the common format and

516
00:20:47,700 --> 00:20:54,419
then send it up um via API

517
00:20:52,080 --> 00:20:55,980
um and then if two different Labs get

518
00:20:54,419 --> 00:20:57,660
their different results then this is

519
00:20:55,980 --> 00:20:58,799
what we show them a diff

520
00:20:57,660 --> 00:21:00,780
um and then they have them basically

521
00:20:58,799 --> 00:21:02,760
they usually have a meeting bring in the

522
00:21:00,780 --> 00:21:05,880
clinicians and the big guns of their

523
00:21:02,760 --> 00:21:07,799
labs to decide what happens

524
00:21:05,880 --> 00:21:09,419
um and then they sort it out go through

525
00:21:07,799 --> 00:21:11,100
a workflow one of them change well

526
00:21:09,419 --> 00:21:13,080
ideally one of them changes to agree

527
00:21:11,100 --> 00:21:14,640
with the other one and then uh it

528
00:21:13,080 --> 00:21:16,679
eventually Cascades through their system

529
00:21:14,640 --> 00:21:19,140
through the API again and then we mark

530
00:21:16,679 --> 00:21:21,059
it as resolved

531
00:21:19,140 --> 00:21:22,919
so yeah one of the troubles

532
00:21:21,059 --> 00:21:26,340
um is how to handle data from many Labs

533
00:21:22,919 --> 00:21:28,860
every state does their own thing and so

534
00:21:26,340 --> 00:21:31,440
we had to write custom converters to put

535
00:21:28,860 --> 00:21:34,940
into Json so sometimes we run it on

536
00:21:31,440 --> 00:21:37,559
their systems sometimes they just dump

537
00:21:34,940 --> 00:21:40,679
data on an S3 bucket and then we run the

538
00:21:37,559 --> 00:21:42,900
code on our cloud system and send it

539
00:21:40,679 --> 00:21:47,520
from our Cloud systems back to another

540
00:21:42,900 --> 00:21:50,039
through the API and yeah so we we allow

541
00:21:47,520 --> 00:21:51,780
basically everything we allow people to

542
00:21:50,039 --> 00:21:53,400
to give every bit of information they

543
00:21:51,780 --> 00:21:54,960
have but after but after we've

544
00:21:53,400 --> 00:21:57,179
investigated and asked for them what it

545
00:21:54,960 --> 00:22:00,179
is then we can sort of add a bit of type

546
00:21:57,179 --> 00:22:01,919
information such as float or int and we

547
00:22:00,179 --> 00:22:04,080
try and get them to use a we use a

548
00:22:01,919 --> 00:22:06,900
controlled vocabulary like an ontology

549
00:22:04,080 --> 00:22:09,059
so if a lab calls it population

550
00:22:06,900 --> 00:22:12,120
frequency another lab calls it pop freak

551
00:22:09,059 --> 00:22:13,679
or something we will convert that and

552
00:22:12,120 --> 00:22:15,720
send it all up on that you know with the

553
00:22:13,679 --> 00:22:18,120
consistency on the API so we try and

554
00:22:15,720 --> 00:22:19,679
hide the um the differences in those

555
00:22:18,120 --> 00:22:22,620
connector programs

556
00:22:19,679 --> 00:22:24,780
so here's what the Json looks like and

557
00:22:22,620 --> 00:22:27,240
the advantage of this is we have I don't

558
00:22:24,780 --> 00:22:29,580
know maybe 100 150 different fields that

559
00:22:27,240 --> 00:22:33,480
we collect from all the different Labs

560
00:22:29,580 --> 00:22:35,940
um and uh what what we do is we've like

561
00:22:33,480 --> 00:22:37,740
um so here there's Nomad allele number

562
00:22:35,940 --> 00:22:39,840
which is how many

563
00:22:37,740 --> 00:22:41,720
um counts this has been seen in in a

564
00:22:39,840 --> 00:22:44,400
hundred thousand person population

565
00:22:41,720 --> 00:22:48,059
survey we know that's going to be an

566
00:22:44,400 --> 00:22:50,580
integer so we have a basically a key

567
00:22:48,059 --> 00:22:52,140
with an integer on it and we then when

568
00:22:50,580 --> 00:22:55,080
we get the data we can run a validation

569
00:22:52,140 --> 00:22:58,140
and say this is supposed to be an INT

570
00:22:55,080 --> 00:22:59,640
um you know uh it's a it's a string and

571
00:22:58,140 --> 00:23:01,260
we can like raise a little warning and

572
00:22:59,640 --> 00:23:03,260
stuff like that which just helps keep

573
00:23:01,260 --> 00:23:07,140
the data pretty clean

574
00:23:03,260 --> 00:23:08,640
so um yeah here's the project timeline

575
00:23:07,140 --> 00:23:11,340
um so basically I started this so I work

576
00:23:08,640 --> 00:23:14,700
as a I work as a researcher on genomics

577
00:23:11,340 --> 00:23:17,820
and I people kept coming to me asking

578
00:23:14,700 --> 00:23:19,320
for me to do filtering and I said why

579
00:23:17,820 --> 00:23:22,080
don't I write some software so that

580
00:23:19,320 --> 00:23:25,919
people don't bother me anymore and then

581
00:23:22,080 --> 00:23:28,020
it got used and then the researchers

582
00:23:25,919 --> 00:23:29,580
started using it then the diagnostic

583
00:23:28,020 --> 00:23:33,780
people started using it we ended up

584
00:23:29,580 --> 00:23:36,960
putting basically we won a um a project

585
00:23:33,780 --> 00:23:39,659
to sort of people have this very rare

586
00:23:36,960 --> 00:23:42,299
disease and a um they basically have

587
00:23:39,659 --> 00:23:43,740
this online site where all researchers

588
00:23:42,299 --> 00:23:45,960
from around the world can upload their

589
00:23:43,740 --> 00:23:47,820
data and share it and we got we're

590
00:23:45,960 --> 00:23:51,120
getting external funding from that from

591
00:23:47,820 --> 00:23:54,240
the states and that's still ongoing that

592
00:23:51,120 --> 00:23:56,760
was um 2018 the first funding and then

593
00:23:54,240 --> 00:24:00,120
um we hired then we won the share uh so

594
00:23:56,760 --> 00:24:01,559
sharing was a project that we had that

595
00:24:00,120 --> 00:24:04,380
Australian genomics wanted to do and

596
00:24:01,559 --> 00:24:06,659
variant grid 1 the tender for that and

597
00:24:04,380 --> 00:24:07,500
we used thanks um yeah the tender for

598
00:24:06,659 --> 00:24:10,140
that

599
00:24:07,500 --> 00:24:12,000
um and use variant grid technology as

600
00:24:10,140 --> 00:24:14,100
the base for sharing it so we've sort of

601
00:24:12,000 --> 00:24:16,440
have the ability to skin it and turn on

602
00:24:14,100 --> 00:24:19,620
and off things through settings to make

603
00:24:16,440 --> 00:24:22,200
the site work however the client wants

604
00:24:19,620 --> 00:24:23,340
it and you know swap the CSS out and

605
00:24:22,200 --> 00:24:25,260
stuff

606
00:24:23,340 --> 00:24:27,720
um yeah so in 2019 we've got a second

607
00:24:25,260 --> 00:24:30,299
developer which was um amazing

608
00:24:27,720 --> 00:24:31,980
um and 2020 we went open source and you

609
00:24:30,299 --> 00:24:34,500
can go to the GitHub page and check it

610
00:24:31,980 --> 00:24:36,480
out I've um in science

611
00:24:34,500 --> 00:24:39,679
um you're supposed to write a paper to

612
00:24:36,480 --> 00:24:42,059
advertise your project and I've been um

613
00:24:39,679 --> 00:24:43,980
I don't know I'm a programmer not a

614
00:24:42,059 --> 00:24:45,720
writer and I've my boss keeps on my back

615
00:24:43,980 --> 00:24:46,919
to write a paper and one day I'll get to

616
00:24:45,720 --> 00:24:49,500
it

617
00:24:46,919 --> 00:24:51,059
um but yeah and then so it got

618
00:24:49,500 --> 00:24:52,260
diagnostic use and it's been used by

619
00:24:51,059 --> 00:24:55,440
sapath

620
00:24:52,260 --> 00:24:57,000
um and yeah so that's how it's going so

621
00:24:55,440 --> 00:24:58,260
thanks a lot

622
00:24:57,000 --> 00:24:58,660
um cheers

623
00:24:58,260 --> 00:24:59,700
[Applause]

624
00:24:58,660 --> 00:25:06,890
[Music]

625
00:24:59,700 --> 00:25:06,890
[Applause]

626
00:25:10,500 --> 00:25:16,679
all right thank you very much David what

627
00:25:13,320 --> 00:25:18,220
a great uh sequence of slides can we

628
00:25:16,679 --> 00:25:18,540
have a round of applause for David

629
00:25:18,220 --> 00:25:21,650
[Music]

630
00:25:18,540 --> 00:25:21,650
[Applause]

631
00:25:23,340 --> 00:25:29,039
now we do have a few minutes for

632
00:25:26,640 --> 00:25:30,900
questions so if anybody in the room

633
00:25:29,039 --> 00:25:32,580
wants to put their hand up or if anybody

634
00:25:30,900 --> 00:25:35,580
wants to type their questions into

635
00:25:32,580 --> 00:25:37,740
Discord and we can read them out here

636
00:25:35,580 --> 00:25:39,480
please do so

637
00:25:37,740 --> 00:25:41,159
I'll give you a moment to type things

638
00:25:39,480 --> 00:25:43,860
into Discord because I know it can be

639
00:25:41,159 --> 00:25:46,860
hard to raise your hands sometimes and

640
00:25:43,860 --> 00:25:48,779
I'll check and answer them later if um

641
00:25:46,860 --> 00:25:52,640
anyone else say anything else so but

642
00:25:48,779 --> 00:25:52,640
yeah feel free to ask any questions

643
00:25:52,679 --> 00:25:58,100
all right we have a question from down

644
00:25:54,960 --> 00:25:58,100
here near the front

645
00:25:59,520 --> 00:26:04,080
um yeah I guess I was just wondering how

646
00:26:01,440 --> 00:26:06,559
kind of easy and usable you find the

647
00:26:04,080 --> 00:26:10,919
kind of ecosystem of

648
00:26:06,559 --> 00:26:12,900
data apis like that ncbi provides for

649
00:26:10,919 --> 00:26:15,539
looking up your sequences and stuff

650
00:26:12,900 --> 00:26:16,980
because I've made my way slowly into

651
00:26:15,539 --> 00:26:20,039
this world and yeah I found it a

652
00:26:16,980 --> 00:26:20,940
struggle yeah um okay that's a great

653
00:26:20,039 --> 00:26:22,620
question

654
00:26:20,940 --> 00:26:24,659
um there are tools there's like um

655
00:26:22,620 --> 00:26:27,240
entrees um there's a python library

656
00:26:24,659 --> 00:26:29,460
biopython and some that sort of manages

657
00:26:27,240 --> 00:26:31,559
that we use that for things like um

658
00:26:29,460 --> 00:26:34,140
querying say how many

659
00:26:31,559 --> 00:26:36,480
um Publications there are for a gene or

660
00:26:34,140 --> 00:26:39,600
you know when was the last time or or

661
00:26:36,480 --> 00:26:42,059
other things but basically the

662
00:26:39,600 --> 00:26:43,260
um yeah the tools aren't

663
00:26:42,059 --> 00:26:46,620
um awesome

664
00:26:43,260 --> 00:26:48,000
um we write yeah basically we um we

665
00:26:46,620 --> 00:26:48,659
always wrap our

666
00:26:48,000 --> 00:26:51,659
um

667
00:26:48,659 --> 00:26:53,640
API calls in uh with timeouts and

668
00:26:51,659 --> 00:26:55,980
exceptions because the services are up

669
00:26:53,640 --> 00:26:57,840
and down uh I don't know it's um but

670
00:26:55,980 --> 00:26:58,500
yeah it's a bit of a weird

671
00:26:57,840 --> 00:27:01,260
um

672
00:26:58,500 --> 00:27:03,240
yeah basically a lot of work was done a

673
00:27:01,260 --> 00:27:05,220
long time ago um by

674
00:27:03,240 --> 00:27:06,659
um people that yeah there's a lot of

675
00:27:05,220 --> 00:27:08,220
some somewhat questionable decisions

676
00:27:06,659 --> 00:27:11,820
that you wouldn't necessarily do it this

677
00:27:08,220 --> 00:27:13,380
way uh now and um but yeah it's um it's

678
00:27:11,820 --> 00:27:15,179
a big yeah that's why we need more

679
00:27:13,380 --> 00:27:17,100
programmers in the field

680
00:27:15,179 --> 00:27:20,100
um get get our sort our apis out and

681
00:27:17,100 --> 00:27:23,520
yeah there's not there's a lot of um uh

682
00:27:20,100 --> 00:27:25,620
custom file formats and wacky old um you

683
00:27:23,520 --> 00:27:27,179
know really inefficient old file formats

684
00:27:25,620 --> 00:27:29,700
that are there instead of you know in my

685
00:27:27,179 --> 00:27:32,880
in my ideal world every service would

686
00:27:29,700 --> 00:27:34,679
have a rest API and it would be trivial

687
00:27:32,880 --> 00:27:36,960
to get everything but that's not a world

688
00:27:34,679 --> 00:27:38,340
we live in yet so but yeah more

689
00:27:36,960 --> 00:27:40,260
programmers would would make that world

690
00:27:38,340 --> 00:27:42,539
a reality so

691
00:27:40,260 --> 00:27:43,620
consider bioinformatics

692
00:27:42,539 --> 00:27:45,960
yeah

693
00:27:43,620 --> 00:27:49,559
all right we have one question from the

694
00:27:45,960 --> 00:27:52,140
Discord uh do the findings push to or

695
00:27:49,559 --> 00:27:55,440
pull from open targets genetics

696
00:27:52,140 --> 00:27:58,200
open targets genetics

697
00:27:55,440 --> 00:27:59,820
um I'm not sure so we go to clinvar as

698
00:27:58,200 --> 00:28:01,799
well which is a very common oh that's

699
00:27:59,820 --> 00:28:04,620
the most popular as far as I know

700
00:28:01,799 --> 00:28:07,559
um variant classification sharing uh

701
00:28:04,620 --> 00:28:10,340
platform and we collect a lot more data

702
00:28:07,559 --> 00:28:12,480
than they do and we also

703
00:28:10,340 --> 00:28:14,640
collect stuff that hasn't sort of been

704
00:28:12,480 --> 00:28:17,100
approved for public use so it's behind a

705
00:28:14,640 --> 00:28:20,760
like a private server and allows the um

706
00:28:17,100 --> 00:28:23,640
uh sort of labs across the country to

707
00:28:20,760 --> 00:28:27,120
sort out any problems out of

708
00:28:23,640 --> 00:28:28,799
um you know internally and then after a

709
00:28:27,120 --> 00:28:31,020
while they've sort everything out and

710
00:28:28,799 --> 00:28:32,640
then they share it and then we do once

711
00:28:31,020 --> 00:28:34,620
we sort of set things to the public

712
00:28:32,640 --> 00:28:36,720
share level then we go through and

713
00:28:34,620 --> 00:28:39,779
collect everything and send it to the

714
00:28:36,720 --> 00:28:41,640
external databases for everyone else to

715
00:28:39,779 --> 00:28:44,520
be able to look at so we do talk to

716
00:28:41,640 --> 00:28:45,900
other databases and push it out and we

717
00:28:44,520 --> 00:28:47,640
also pull those in so we can do

718
00:28:45,900 --> 00:28:49,980
comparisons if if ours are different

719
00:28:47,640 --> 00:28:52,320
from theirs so we both push and pull to

720
00:28:49,980 --> 00:28:54,360
external databases

721
00:28:52,320 --> 00:28:56,400
all right that's all the question all

722
00:28:54,360 --> 00:28:59,100
the time we have for questions uh I'm

723
00:28:56,400 --> 00:29:00,720
have the traditional speakers gift for

724
00:28:59,100 --> 00:29:02,820
David

725
00:29:00,720 --> 00:29:06,299
the wonderful

726
00:29:02,820 --> 00:29:09,260
pycon a youth speakers mug thank you one

727
00:29:06,299 --> 00:29:09,260
more round of applause for David