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

2
00:00:08,400 --> 00:00:12,120
thank you um I'd like to thank the

3
00:00:10,620 --> 00:00:13,679
organizers for giving me the opportunity

4
00:00:12,120 --> 00:00:15,299
to talk to you today

5
00:00:13,679 --> 00:00:17,340
um you know this is my first time

6
00:00:15,299 --> 00:00:19,140
presenting at pycon I usually present to

7
00:00:17,340 --> 00:00:20,460
rooms of scientific researchers is a

8
00:00:19,140 --> 00:00:22,500
very different style of talk than what

9
00:00:20,460 --> 00:00:24,000
I'm used to but I'm really excited to

10
00:00:22,500 --> 00:00:25,980
tell you the story of an ongoing

11
00:00:24,000 --> 00:00:28,140
academic open source project to build a

12
00:00:25,980 --> 00:00:29,820
biological database for experimental

13
00:00:28,140 --> 00:00:31,619
data that's generated by researchers and

14
00:00:29,820 --> 00:00:33,120
genomics and I'm really excited to share

15
00:00:31,619 --> 00:00:34,860
it with this community because Django

16
00:00:33,120 --> 00:00:36,120
really plays a starring role and while

17
00:00:34,860 --> 00:00:37,640
there's a bit of technical stuff and

18
00:00:36,120 --> 00:00:40,140
I'll explain a little bit of the biology

19
00:00:37,640 --> 00:00:42,120
to kind of motivate things this is

20
00:00:40,140 --> 00:00:44,700
mostly a talk about how we try to match

21
00:00:42,120 --> 00:00:46,620
our team shifting expertise and variable

22
00:00:44,700 --> 00:00:49,140
level of resourcing to available open

23
00:00:46,620 --> 00:00:50,820
source Tools in order to achieve you

24
00:00:49,140 --> 00:00:52,260
know the goals and requirements of this

25
00:00:50,820 --> 00:00:54,180
project that really changed a lot over

26
00:00:52,260 --> 00:00:55,559
time and I hope it also maybe gives you

27
00:00:54,180 --> 00:00:57,300
a little bit of insight into what it's

28
00:00:55,559 --> 00:00:59,280
like to do software development in an

29
00:00:57,300 --> 00:01:01,320
academic research context which I think

30
00:00:59,280 --> 00:01:03,239
is both in some ways similar and very

31
00:01:01,320 --> 00:01:04,440
different to how many people do software

32
00:01:03,239 --> 00:01:07,500
development

33
00:01:04,440 --> 00:01:10,080
so a little bit about me so I am a

34
00:01:07,500 --> 00:01:11,760
computational biologist I'm a researcher

35
00:01:10,080 --> 00:01:13,680
at the Walton Eliza Hall Institute of

36
00:01:11,760 --> 00:01:15,720
Medical Research or wihai in Melbourne

37
00:01:13,680 --> 00:01:17,340
Australia this work is all being been

38
00:01:15,720 --> 00:01:18,479
done in close collaboration with

39
00:01:17,340 --> 00:01:19,680
colleagues at the University of

40
00:01:18,479 --> 00:01:21,900
Washington and the broadman media

41
00:01:19,680 --> 00:01:24,000
Institute for precision medicine in

42
00:01:21,900 --> 00:01:26,759
Seattle and the work that I'm talking

43
00:01:24,000 --> 00:01:28,140
about has been supported by competitive

44
00:01:26,759 --> 00:01:30,659
grants from the medical research future

45
00:01:28,140 --> 00:01:32,280
fund here in Australia and also from the

46
00:01:30,659 --> 00:01:33,840
national Human Genome Research Institute

47
00:01:32,280 --> 00:01:36,180
which is part of the National Institutes

48
00:01:33,840 --> 00:01:37,860
of Health in the United States

49
00:01:36,180 --> 00:01:40,020
and so with that I'd like to just tell

50
00:01:37,860 --> 00:01:43,640
you a little bit about what kind of data

51
00:01:40,020 --> 00:01:43,640
we're storing and why we're doing that

52
00:01:44,400 --> 00:01:48,420
so fundamentally the data that we're

53
00:01:46,259 --> 00:01:50,640
storing is about protein so proteins are

54
00:01:48,420 --> 00:01:52,200
a type of biological molecule they have

55
00:01:50,640 --> 00:01:53,280
a lot of different functions there you

56
00:01:52,200 --> 00:01:55,799
can think of them sort of as Little

57
00:01:53,280 --> 00:01:58,560
Machines inside your cells that perform

58
00:01:55,799 --> 00:02:01,500
all of the functions that your cells

59
00:01:58,560 --> 00:02:02,820
need to perform so antibodies are a kind

60
00:02:01,500 --> 00:02:04,439
of protein that I think everyone's

61
00:02:02,820 --> 00:02:06,860
learned a lot about over the last few

62
00:02:04,439 --> 00:02:09,179
years that are involved in immunity

63
00:02:06,860 --> 00:02:10,319
enzymes are a class of proteins that

64
00:02:09,179 --> 00:02:11,700
assist in all types of different

65
00:02:10,319 --> 00:02:14,099
chemical reactions whether that's

66
00:02:11,700 --> 00:02:15,660
metabolism or synthesizing the various

67
00:02:14,099 --> 00:02:17,580
different chemicals that your body needs

68
00:02:15,660 --> 00:02:19,620
they have a lot of different different

69
00:02:17,580 --> 00:02:21,660
roles and then there are a lot of other

70
00:02:19,620 --> 00:02:23,520
different sort of

71
00:02:21,660 --> 00:02:25,739
different functions that proteins have

72
00:02:23,520 --> 00:02:27,660
like forming the physical structure of

73
00:02:25,739 --> 00:02:30,180
your cells to things like this chaperone

74
00:02:27,660 --> 00:02:32,640
protein here which actually forms like a

75
00:02:30,180 --> 00:02:34,680
little canister that can open up and

76
00:02:32,640 --> 00:02:37,500
contain other proteins and help them

77
00:02:34,680 --> 00:02:39,599
stabilize and so and I'll just mention

78
00:02:37,500 --> 00:02:41,400
that all of these images of these

79
00:02:39,599 --> 00:02:42,900
protein structures that I'm showing are

80
00:02:41,400 --> 00:02:44,819
from the excellent molecule of the month

81
00:02:42,900 --> 00:02:47,180
series which is run by the protein Data

82
00:02:44,819 --> 00:02:49,560
Bank sort of a kind of a public science

83
00:02:47,180 --> 00:02:51,120
sort of series which I highly recommend

84
00:02:49,560 --> 00:02:52,620
if you want to learn more about the

85
00:02:51,120 --> 00:02:56,420
interesting things that proteins can do

86
00:02:52,620 --> 00:02:56,420
and see more of these cool images

87
00:02:57,599 --> 00:03:02,160
now Decades of work in molecular biology

88
00:03:00,060 --> 00:03:05,640
has taught us these relationship which

89
00:03:02,160 --> 00:03:08,459
is that DNA gets transcribed into RNA

90
00:03:05,640 --> 00:03:10,379
and RNA gets translated into protein and

91
00:03:08,459 --> 00:03:11,819
it's helped us work out the mechanics of

92
00:03:10,379 --> 00:03:13,319
this and the grammar and I'm not going

93
00:03:11,819 --> 00:03:15,060
to go into this because I don't want to

94
00:03:13,319 --> 00:03:16,500
spend all of this time giving you a

95
00:03:15,060 --> 00:03:17,640
public science lecture although if you

96
00:03:16,500 --> 00:03:19,140
want to learn more about any of this

97
00:03:17,640 --> 00:03:20,760
stuff you can come Corner me at the con

98
00:03:19,140 --> 00:03:22,920
at the conference later and I'll answer

99
00:03:20,760 --> 00:03:24,420
any of your questions so we're not going

100
00:03:22,920 --> 00:03:26,400
to focus on this but the takeaway here

101
00:03:24,420 --> 00:03:28,440
is that the DNA sequence which is stored

102
00:03:26,400 --> 00:03:30,599
in your genome inside the nucleus of

103
00:03:28,440 --> 00:03:32,940
your cells contains all of the

104
00:03:30,599 --> 00:03:35,760
information that is needed to determine

105
00:03:32,940 --> 00:03:37,680
the sequence of the protein so the

106
00:03:35,760 --> 00:03:39,840
information flows largely in this way

107
00:03:37,680 --> 00:03:42,120
and that's the kind of the take-home

108
00:03:39,840 --> 00:03:45,299
message here

109
00:03:42,120 --> 00:03:47,580
now proteins are chains of amino acid

110
00:03:45,299 --> 00:03:49,440
it's a biological polymer the amino

111
00:03:47,580 --> 00:03:51,780
acids are the subunits and it's just a

112
00:03:49,440 --> 00:03:53,700
single chain of amino acids there are 20

113
00:03:51,780 --> 00:03:55,739
canonical amino acids so we can give

114
00:03:53,700 --> 00:03:57,060
each of them a single letter and we can

115
00:03:55,739 --> 00:03:59,099
just write the sequence as a string

116
00:03:57,060 --> 00:04:03,319
which was what we're doing here so this

117
00:03:59,099 --> 00:04:06,200
is the the amino acid sequence

118
00:04:03,319 --> 00:04:09,120
of this protein that I showed you before

119
00:04:06,200 --> 00:04:11,700
and so the important thing that I want

120
00:04:09,120 --> 00:04:14,519
to tell you about this is that the

121
00:04:11,700 --> 00:04:16,260
sequence of amino acids determines the

122
00:04:14,519 --> 00:04:19,019
structure of the protein so this this

123
00:04:16,260 --> 00:04:21,060
chain of amino acids is going to fold up

124
00:04:19,019 --> 00:04:22,500
and it's going to fold up into some

125
00:04:21,060 --> 00:04:24,900
shape based on the biophysical

126
00:04:22,500 --> 00:04:26,699
properties of those amino acids and the

127
00:04:24,900 --> 00:04:28,800
shape and the position of the atoms in

128
00:04:26,699 --> 00:04:30,540
that protein structure are what allows

129
00:04:28,800 --> 00:04:32,699
the protein to perform its function or

130
00:04:30,540 --> 00:04:35,340
to do its job so that's what's important

131
00:04:32,699 --> 00:04:37,979
and some of you who you know follow this

132
00:04:35,340 --> 00:04:40,020
uh kind of aspect of science May

133
00:04:37,979 --> 00:04:41,460
recognize this as the protein folding

134
00:04:40,020 --> 00:04:42,660
problem and it's something that a lot of

135
00:04:41,460 --> 00:04:44,820
groups have spent a lot of time trying

136
00:04:42,660 --> 00:04:45,800
to predict is how do you go from a

137
00:04:44,820 --> 00:04:49,199
sequence

138
00:04:45,800 --> 00:04:50,699
to the structure

139
00:04:49,199 --> 00:04:52,560
um and so the structure allows the

140
00:04:50,699 --> 00:04:53,880
protein to do its function so the

141
00:04:52,560 --> 00:04:55,620
function of this protein which is a

142
00:04:53,880 --> 00:04:57,600
caspase protein it's involved in program

143
00:04:55,620 --> 00:04:59,940
cell death is that it kills the cell

144
00:04:57,600 --> 00:05:01,320
when it's time to kill the cell and

145
00:04:59,940 --> 00:05:03,180
there's a lot of context in which that's

146
00:05:01,320 --> 00:05:04,639
important including cancer and also

147
00:05:03,180 --> 00:05:07,500
normal development

148
00:05:04,639 --> 00:05:09,300
but the take home here is that the amino

149
00:05:07,500 --> 00:05:10,979
acid sequence determines whether or not

150
00:05:09,300 --> 00:05:14,520
the protein is going to be able to do

151
00:05:10,979 --> 00:05:16,860
its function based on its structure

152
00:05:14,520 --> 00:05:18,479
now I told you before that we can get to

153
00:05:16,860 --> 00:05:20,100
the amino acid sequence from the DNA

154
00:05:18,479 --> 00:05:21,780
sequence so if I just swap over to the

155
00:05:20,100 --> 00:05:23,759
DNA sequence here we can say that the

156
00:05:21,780 --> 00:05:25,440
DNA sequence is what determines the

157
00:05:23,759 --> 00:05:27,539
function of whether the protein can do

158
00:05:25,440 --> 00:05:30,840
its function and so then a change so if

159
00:05:27,539 --> 00:05:32,520
we change this a to a t which is there

160
00:05:30,840 --> 00:05:34,080
in red kind of in the in the bottom

161
00:05:32,520 --> 00:05:34,800
middle there

162
00:05:34,080 --> 00:05:37,380
um

163
00:05:34,800 --> 00:05:39,539
you know that may cause a change in the

164
00:05:37,380 --> 00:05:41,280
that DNA change may cause a change in

165
00:05:39,539 --> 00:05:43,020
the amino acid sequence that change in

166
00:05:41,280 --> 00:05:44,699
the amino acid sequence may make a

167
00:05:43,020 --> 00:05:46,740
meaningful change to the structure and

168
00:05:44,699 --> 00:05:48,419
that change to the structure may have a

169
00:05:46,740 --> 00:05:50,639
meaningful impact on whether or not the

170
00:05:48,419 --> 00:05:52,680
protein can do its function and despite

171
00:05:50,639 --> 00:05:55,259
the success of tools like Alpha full 2

172
00:05:52,680 --> 00:05:57,180
and primate Ai and other sort of large

173
00:05:55,259 --> 00:05:59,160
predictive models the effect of a

174
00:05:57,180 --> 00:06:01,080
specific amino acid change on a protein

175
00:05:59,160 --> 00:06:02,880
structure and therefore function remains

176
00:06:01,080 --> 00:06:04,320
very difficult to predict and there's

177
00:06:02,880 --> 00:06:06,180
still really no substitute for doing

178
00:06:04,320 --> 00:06:08,160
experiment and advances in genomic

179
00:06:06,180 --> 00:06:10,800
technology particularly High throughput

180
00:06:08,160 --> 00:06:12,840
DNA sequencing and high throughput DNA

181
00:06:10,800 --> 00:06:14,759
synthesis allows us to look at these

182
00:06:12,840 --> 00:06:16,320
changes at scale and that's the type of

183
00:06:14,759 --> 00:06:19,740
experimental data that we're trying to

184
00:06:16,320 --> 00:06:22,139
store in this database

185
00:06:19,740 --> 00:06:24,419
um and so what researchers can do now is

186
00:06:22,139 --> 00:06:26,639
we can measure every change in a single

187
00:06:24,419 --> 00:06:28,560
protein at once so we make a population

188
00:06:26,639 --> 00:06:31,740
of cells where each one has a different

189
00:06:28,560 --> 00:06:33,479
variant of the protein and then we

190
00:06:31,740 --> 00:06:35,400
compete them against each other in some

191
00:06:33,479 --> 00:06:37,979
sort of functional assay and we see

192
00:06:35,400 --> 00:06:40,560
whether or not the protein Works under

193
00:06:37,979 --> 00:06:43,319
that assay condition or not and then

194
00:06:40,560 --> 00:06:44,340
through the application of high

195
00:06:43,319 --> 00:06:46,199
throughput sequencing and then

196
00:06:44,340 --> 00:06:47,759
bioinformatics what we can do is we can

197
00:06:46,199 --> 00:06:50,699
take the data that comes out of this

198
00:06:47,759 --> 00:06:52,500
assay and we can assess a functional

199
00:06:50,699 --> 00:06:54,840
score for each one of those changes that

200
00:06:52,500 --> 00:06:56,460
we're interested in and we can plot it

201
00:06:54,840 --> 00:06:58,020
you know using a heat looking at a heat

202
00:06:56,460 --> 00:07:00,300
map like this and we get all of this

203
00:06:58,020 --> 00:07:02,340
information about which positions in the

204
00:07:00,300 --> 00:07:03,720
protein are important what changes are

205
00:07:02,340 --> 00:07:05,220
tolerated and there's all kinds of

206
00:07:03,720 --> 00:07:06,780
interesting applications for this data

207
00:07:05,220 --> 00:07:08,580
if you want to learn more about the

208
00:07:06,780 --> 00:07:10,500
bioinformatics behind calculating these

209
00:07:08,580 --> 00:07:13,380
scores and how all of this works and

210
00:07:10,500 --> 00:07:15,120
here this stuff explained by a software

211
00:07:13,380 --> 00:07:16,380
engineer rather than by me which I think

212
00:07:15,120 --> 00:07:18,419
is going to give you some interesting

213
00:07:16,380 --> 00:07:22,080
additional context please go see Nick

214
00:07:18,419 --> 00:07:23,759
Morse talk tomorrow which I think is at

215
00:07:22,080 --> 00:07:25,800
noon in this room unless the schedule

216
00:07:23,759 --> 00:07:27,539
has changed

217
00:07:25,800 --> 00:07:29,220
um so that's the background that I

218
00:07:27,539 --> 00:07:31,560
wanted to give you about

219
00:07:29,220 --> 00:07:34,440
um the the science I hope that it was

220
00:07:31,560 --> 00:07:35,880
you know somewhat clear and uh and that

221
00:07:34,440 --> 00:07:38,099
gives you some idea of kind of what

222
00:07:35,880 --> 00:07:40,020
we're dealing with and so we wanted to

223
00:07:38,099 --> 00:07:41,819
build a database to store all this

224
00:07:40,020 --> 00:07:43,740
information and so why do we want a

225
00:07:41,819 --> 00:07:46,020
database when we talk about research

226
00:07:43,740 --> 00:07:48,060
data and um

227
00:07:46,020 --> 00:07:48,840
and and what we want to do with it and

228
00:07:48,060 --> 00:07:51,120
and

229
00:07:48,840 --> 00:07:52,500
how to have it have the most impact a

230
00:07:51,120 --> 00:07:55,199
lot of times we talk about how we want

231
00:07:52,500 --> 00:07:57,180
data to be fair and fair data is an

232
00:07:55,199 --> 00:07:59,639
acronym for findable accessible

233
00:07:57,180 --> 00:08:01,319
interoperable and reusable and so we

234
00:07:59,639 --> 00:08:03,419
want all the public data that's being

235
00:08:01,319 --> 00:08:05,099
you know generated especially the stuff

236
00:08:03,419 --> 00:08:06,599
that's being funded by government grants

237
00:08:05,099 --> 00:08:08,880
and is generated with public funds we

238
00:08:06,599 --> 00:08:10,560
want it all all that data to be fair we

239
00:08:08,880 --> 00:08:12,240
want it to be open we want everybody to

240
00:08:10,560 --> 00:08:14,819
be able to use it and so we need to

241
00:08:12,240 --> 00:08:16,919
organize it in some way

242
00:08:14,819 --> 00:08:18,780
and uh one of the best ways to do this

243
00:08:16,919 --> 00:08:19,919
is through applying data standards so

244
00:08:18,780 --> 00:08:21,780
I'm just going to show you some quick

245
00:08:19,919 --> 00:08:24,060
examples of what actual data looks like

246
00:08:21,780 --> 00:08:25,979
from actual papers chosen basically at

247
00:08:24,060 --> 00:08:27,660
random by what I just had downloaded on

248
00:08:25,979 --> 00:08:29,940
my laptop and that I opened up in the

249
00:08:27,660 --> 00:08:31,500
spreadsheet program so here's one where

250
00:08:29,940 --> 00:08:33,120
you actually have the amino acid

251
00:08:31,500 --> 00:08:35,039
sequences for all the different variants

252
00:08:33,120 --> 00:08:36,839
listed there are some quantities that

253
00:08:35,039 --> 00:08:39,120
were measured here's another one that

254
00:08:36,839 --> 00:08:42,779
uses a very sort of terse single letter

255
00:08:39,120 --> 00:08:44,760
position single letter code to describe

256
00:08:42,779 --> 00:08:47,640
the variations on a sequence which is

257
00:08:44,760 --> 00:08:49,019
not depicted in the spreadsheet here's

258
00:08:47,640 --> 00:08:50,160
another one that does something similar

259
00:08:49,019 --> 00:08:52,140
they measured a whole bunch of

260
00:08:50,160 --> 00:08:53,880
additional different things additional

261
00:08:52,140 --> 00:08:56,040
stuff during their assay it's not

262
00:08:53,880 --> 00:08:58,320
necessarily clear from this screenshot

263
00:08:56,040 --> 00:09:00,060
what the most important value is if you

264
00:08:58,320 --> 00:09:01,860
just want to know what is this protein

265
00:09:00,060 --> 00:09:02,519
variant do

266
00:09:01,860 --> 00:09:04,620
um

267
00:09:02,519 --> 00:09:06,420
and here's another another different one

268
00:09:04,620 --> 00:09:08,459
that is actually

269
00:09:06,420 --> 00:09:09,899
represented as a matrix as you would

270
00:09:08,459 --> 00:09:11,459
make if you wanted to build one of those

271
00:09:09,899 --> 00:09:13,140
heat maps that I showed you briefly on

272
00:09:11,459 --> 00:09:15,660
the previous slide so this is kind of

273
00:09:13,140 --> 00:09:17,640
all over the place right and if what you

274
00:09:15,660 --> 00:09:19,620
want to do is download a data set

275
00:09:17,640 --> 00:09:21,779
because you're interested in some

276
00:09:19,620 --> 00:09:23,399
protein and what it does

277
00:09:21,779 --> 00:09:25,440
um you're going to maybe have a lot of

278
00:09:23,399 --> 00:09:27,360
kind of munging and data cleaning and

279
00:09:25,440 --> 00:09:29,160
you know parsing that you have to do for

280
00:09:27,360 --> 00:09:30,839
this and even worse if you want to do

281
00:09:29,160 --> 00:09:32,220
something like you know do a big

282
00:09:30,839 --> 00:09:34,140
meta-analysis or build a machine

283
00:09:32,220 --> 00:09:37,800
learning model that is a huge amount of

284
00:09:34,140 --> 00:09:40,080
work for you or probably your student to

285
00:09:37,800 --> 00:09:42,240
kind of get all of this stuff into shape

286
00:09:40,080 --> 00:09:45,240
so that you can actually start doing the

287
00:09:42,240 --> 00:09:47,700
work so if we can build a database then

288
00:09:45,240 --> 00:09:49,800
store everything in in a standard format

289
00:09:47,700 --> 00:09:52,140
make that available it's really going to

290
00:09:49,800 --> 00:09:54,660
save a lot of people a lot of time

291
00:09:52,140 --> 00:09:56,220
so I'm going to keep coming back to this

292
00:09:54,660 --> 00:09:58,140
project timeline where I'm going to kind

293
00:09:56,220 --> 00:09:59,880
of walk you through like the kind of how

294
00:09:58,140 --> 00:10:01,860
our project went

295
00:09:59,880 --> 00:10:03,420
um and so faced with this increasingly

296
00:10:01,860 --> 00:10:05,760
unmanageable situation in the literature

297
00:10:03,420 --> 00:10:07,019
that I just gave you a sneak peek of

298
00:10:05,760 --> 00:10:08,459
um my colleagues at the University of

299
00:10:07,019 --> 00:10:10,080
Washington met with some of the other

300
00:10:08,459 --> 00:10:11,519
Pioneers in this field and they decided

301
00:10:10,080 --> 00:10:12,720
that our field needed its own database

302
00:10:11,519 --> 00:10:15,540
because the data is sufficiently

303
00:10:12,720 --> 00:10:16,860
different from other types of data to

304
00:10:15,540 --> 00:10:19,100
not be well served by anything that

305
00:10:16,860 --> 00:10:21,240
already existed and at the time

306
00:10:19,100 --> 00:10:23,160
databases were really either kind of

307
00:10:21,240 --> 00:10:25,140
tiny Boutique efforts like you know

308
00:10:23,160 --> 00:10:27,839
information just about a very specific

309
00:10:25,140 --> 00:10:29,279
gene or specific biological process or

310
00:10:27,839 --> 00:10:30,660
they're run by centrally funded

311
00:10:29,279 --> 00:10:32,700
Government research organizations so

312
00:10:30,660 --> 00:10:34,440
building a database that was sort of you

313
00:10:32,700 --> 00:10:37,560
know mid-sized to serve this growing

314
00:10:34,440 --> 00:10:39,660
field seemed really fanciful

315
00:10:37,560 --> 00:10:41,820
and then I attended pycon Au for the

316
00:10:39,660 --> 00:10:43,440
first time in Melbourne in 2016 and I

317
00:10:41,820 --> 00:10:44,940
learned about Django and after some

318
00:10:43,440 --> 00:10:46,920
conversations with some people who are

319
00:10:44,940 --> 00:10:48,420
here in this room today I was filled

320
00:10:46,920 --> 00:10:50,519
with hubris and I contacted my

321
00:10:48,420 --> 00:10:51,959
collaborators in Seattle and I told them

322
00:10:50,519 --> 00:10:54,380
that I could build the database that

323
00:10:51,959 --> 00:10:54,380
they needed

324
00:10:55,190 --> 00:10:58,380
[Applause]

325
00:10:56,399 --> 00:11:00,000
so what did we think we what do we want

326
00:10:58,380 --> 00:11:01,740
to do like what did success look like to

327
00:11:00,000 --> 00:11:03,779
us when we started this project

328
00:11:01,740 --> 00:11:06,180
well we wanted to build a database where

329
00:11:03,779 --> 00:11:08,880
users could upload their data sets

330
00:11:06,180 --> 00:11:10,620
we wanted to build and we wanted to make

331
00:11:08,880 --> 00:11:11,839
it so that users could find and download

332
00:11:10,620 --> 00:11:13,980
data sets that were interested

333
00:11:11,839 --> 00:11:15,360
interesting to them and this is entirely

334
00:11:13,980 --> 00:11:16,500
targeted at researchers who are

335
00:11:15,360 --> 00:11:18,720
generating the data sets that we're

336
00:11:16,500 --> 00:11:22,380
talking about and we needed to make sure

337
00:11:18,720 --> 00:11:24,240
that it could be managed by a very small

338
00:11:22,380 --> 00:11:26,279
team with limited expertise because

339
00:11:24,240 --> 00:11:28,440
that's what we had

340
00:11:26,279 --> 00:11:29,940
so when we got started we got a very

341
00:11:28,440 --> 00:11:31,380
small amount of money with from a

342
00:11:29,940 --> 00:11:33,660
collaborator who had some flexible

343
00:11:31,380 --> 00:11:36,079
funding for which this work was in scope

344
00:11:33,660 --> 00:11:39,360
and we got started and it was just me

345
00:11:36,079 --> 00:11:40,800
and a student who has just finished his

346
00:11:39,360 --> 00:11:42,959
Master's in bioinformatics he wasn't

347
00:11:40,800 --> 00:11:44,399
sure if he wanted to do a PhD and he

348
00:11:42,959 --> 00:11:46,380
decided he wanted to do a little bit of

349
00:11:44,399 --> 00:11:48,360
work first as like you know doing some

350
00:11:46,380 --> 00:11:51,240
software stuff and so he'd done a little

351
00:11:48,360 --> 00:11:52,980
bit of work with like websites and we've

352
00:11:51,240 --> 00:11:55,500
both heard of Django so we figured that

353
00:11:52,980 --> 00:11:57,420
was good enough to get started and in

354
00:11:55,500 --> 00:11:59,100
the beginning we basically followed the

355
00:11:57,420 --> 00:12:01,380
Django girls tutorial but instead of

356
00:11:59,100 --> 00:12:04,339
making a Blog we built a database for

357
00:12:01,380 --> 00:12:04,339
genomic researchers

358
00:12:05,160 --> 00:12:09,300
and one of the really great benefits of

359
00:12:07,620 --> 00:12:10,440
working with Django you know this worked

360
00:12:09,300 --> 00:12:11,760
one of the benefits of working with

361
00:12:10,440 --> 00:12:12,779
Django especially at the beginning of

362
00:12:11,760 --> 00:12:15,959
the project when we didn't know anything

363
00:12:12,779 --> 00:12:17,519
was that it let us really focus on the

364
00:12:15,959 --> 00:12:19,680
higher level Concepts like getting the

365
00:12:17,519 --> 00:12:21,060
organization of the data right

366
00:12:19,680 --> 00:12:23,160
um so we settled on this hierarchical

367
00:12:21,060 --> 00:12:24,779
organization with these nested records

368
00:12:23,160 --> 00:12:27,300
of different types where an experiment

369
00:12:24,779 --> 00:12:29,820
set encapsulates all of the studies in a

370
00:12:27,300 --> 00:12:32,399
single Paper an experiment describes all

371
00:12:29,820 --> 00:12:34,140
the wet lab steps and the sequencing and

372
00:12:32,399 --> 00:12:35,880
then these score set records store the

373
00:12:34,140 --> 00:12:38,519
actual data and how the data analysis

374
00:12:35,880 --> 00:12:40,320
was performed and the each so we kind of

375
00:12:38,519 --> 00:12:41,579
have this hierarchy which becomes

376
00:12:40,320 --> 00:12:43,620
important when we have a more

377
00:12:41,579 --> 00:12:46,560
complicated experimental design so this

378
00:12:43,620 --> 00:12:47,339
is a real example where we had one set

379
00:12:46,560 --> 00:12:49,260
of

380
00:12:47,339 --> 00:12:51,899
one set of protein variants that was

381
00:12:49,260 --> 00:12:54,480
assayed in a bacteriophage model system

382
00:12:51,899 --> 00:12:56,579
and in a yeast model system and then

383
00:12:54,480 --> 00:12:58,620
they were each analyzed to produce two

384
00:12:56,579 --> 00:13:01,440
different data sets

385
00:12:58,620 --> 00:13:03,120
um and uh and the important thing for us

386
00:13:01,440 --> 00:13:04,800
here was that if somebody

387
00:13:03,120 --> 00:13:07,620
grabbed one of these data sets they

388
00:13:04,800 --> 00:13:09,180
would now be able to easily find all of

389
00:13:07,620 --> 00:13:11,399
the other data sets that were related in

390
00:13:09,180 --> 00:13:14,220
the database and they would be able to

391
00:13:11,399 --> 00:13:15,720
see if the raw data had been analyzed in

392
00:13:14,220 --> 00:13:17,940
a different way or there were alternate

393
00:13:15,720 --> 00:13:19,079
representations of the data and so these

394
00:13:17,940 --> 00:13:20,700
were the kinds of things that we could

395
00:13:19,079 --> 00:13:22,500
spend our time thinking about rather

396
00:13:20,700 --> 00:13:25,560
than thinking about how to just get

397
00:13:22,500 --> 00:13:27,060
words to appear in the browser

398
00:13:25,560 --> 00:13:28,800
um

399
00:13:27,060 --> 00:13:30,300
so when we kicked off

400
00:13:28,800 --> 00:13:31,920
um this is kind of the tech stack that

401
00:13:30,300 --> 00:13:34,760
we were working with so that the

402
00:13:31,920 --> 00:13:37,079
database was hosted in a university

403
00:13:34,760 --> 00:13:39,600
Department within a medical school

404
00:13:37,079 --> 00:13:42,540
they're very concerned about making sure

405
00:13:39,600 --> 00:13:43,440
that everything is working

406
00:13:42,540 --> 00:13:46,560
um

407
00:13:43,440 --> 00:13:49,019
and so you know they were still running

408
00:13:46,560 --> 00:13:50,820
on on Centos 6 so we actually had to get

409
00:13:49,019 --> 00:13:52,620
set up with an AWS virtual machine

410
00:13:50,820 --> 00:13:54,720
running Centos 6 so that we could get

411
00:13:52,620 --> 00:13:56,339
that environment set up and then we

412
00:13:54,720 --> 00:13:58,019
could provide an installation procedure

413
00:13:56,339 --> 00:14:00,480
to the IT department so that they could

414
00:13:58,019 --> 00:14:03,959
get all of our stuff working we were

415
00:14:00,480 --> 00:14:05,519
running uh postgres 9.6 for for the back

416
00:14:03,959 --> 00:14:07,560
end because we ended up using Json

417
00:14:05,519 --> 00:14:09,420
Fields a lot asked me about how that

418
00:14:07,560 --> 00:14:12,839
went it's actually it's awesome it's a

419
00:14:09,420 --> 00:14:14,339
success story The IT department asked us

420
00:14:12,839 --> 00:14:16,019
if we could please just use system

421
00:14:14,339 --> 00:14:18,720
python rather than installing something

422
00:14:16,019 --> 00:14:21,360
so that meant Python 3.4 and we were

423
00:14:18,720 --> 00:14:23,700
using Django one we're also running

424
00:14:21,360 --> 00:14:25,740
rabbitmq and celery so we had a task

425
00:14:23,700 --> 00:14:27,779
queue to offload the longer running data

426
00:14:25,740 --> 00:14:29,220
processing and validation tasks when we

427
00:14:27,779 --> 00:14:31,200
were accepting the user data when we're

428
00:14:29,220 --> 00:14:32,399
accepting the uploads

429
00:14:31,200 --> 00:14:33,720
um and I don't want to dwell on this too

430
00:14:32,399 --> 00:14:35,220
much but this was really hard and it was

431
00:14:33,720 --> 00:14:36,660
a lot of work like I have a PhD in

432
00:14:35,220 --> 00:14:39,240
genomics like I didn't even really know

433
00:14:36,660 --> 00:14:41,760
what devops was and here we are like

434
00:14:39,240 --> 00:14:44,040
trying to do this but you know we we did

435
00:14:41,760 --> 00:14:46,500
it we did the thing and so this is what

436
00:14:44,040 --> 00:14:49,199
version 1.0 looked like

437
00:14:46,500 --> 00:14:51,240
um so here we have uh you know a data

438
00:14:49,199 --> 00:14:54,180
table that was based on on data that was

439
00:14:51,240 --> 00:14:56,160
uploaded by the user

440
00:14:54,180 --> 00:14:57,779
um we had automatically generated

441
00:14:56,160 --> 00:14:59,100
accession numbers you know we have like

442
00:14:57,779 --> 00:15:00,480
a title and abstract and some other

443
00:14:59,100 --> 00:15:03,420
stuff that you can see some other you

444
00:15:00,480 --> 00:15:05,579
know textual metadata uh we had linkouts

445
00:15:03,420 --> 00:15:07,139
to uh an external visualization tool

446
00:15:05,579 --> 00:15:08,639
that would make those heat Maps like I

447
00:15:07,139 --> 00:15:11,279
showed you with some extra annotations

448
00:15:08,639 --> 00:15:13,560
and that could download the data from

449
00:15:11,279 --> 00:15:15,180
the server produce those images and then

450
00:15:13,560 --> 00:15:17,639
give you the sort of interactive plot

451
00:15:15,180 --> 00:15:20,459
and that was located posted um by

452
00:15:17,639 --> 00:15:22,380
another collaborating Institution

453
00:15:20,459 --> 00:15:24,300
um we have download wings so that people

454
00:15:22,380 --> 00:15:26,579
could get the data tables out and then

455
00:15:24,300 --> 00:15:28,199
and we have you know data licensing that

456
00:15:26,579 --> 00:15:29,940
was selectable by the user so this kind

457
00:15:28,199 --> 00:15:32,279
of had all of the pieces that we needed

458
00:15:29,940 --> 00:15:34,620
like this is what we wanted to build and

459
00:15:32,279 --> 00:15:36,959
we were able to do it and it took us

460
00:15:34,620 --> 00:15:41,579
about a year so we went from that first

461
00:15:36,959 --> 00:15:44,160
commit to uh in June of 2017 to uh the

462
00:15:41,579 --> 00:15:46,019
first deployment of the full thing we

463
00:15:44,160 --> 00:15:47,639
got a couple little more pots of money

464
00:15:46,019 --> 00:15:49,680
to keep things going because this is an

465
00:15:47,639 --> 00:15:51,779
academic project we wrote a paper we put

466
00:15:49,680 --> 00:15:55,560
it on the pre-print server and then the

467
00:15:51,779 --> 00:15:59,579
paper got published in November of 2019.

468
00:15:55,560 --> 00:16:01,019
and everybody lives happily ever after

469
00:15:59,579 --> 00:16:02,220
um so here's what we thought was going

470
00:16:01,019 --> 00:16:04,820
to happen so we thought we put the

471
00:16:02,220 --> 00:16:04,820
database up

472
00:16:05,639 --> 00:16:08,160
um

473
00:16:06,180 --> 00:16:09,600
and then users would come and they would

474
00:16:08,160 --> 00:16:12,839
upload the data

475
00:16:09,600 --> 00:16:14,220
and then we would you know maybe do some

476
00:16:12,839 --> 00:16:15,120
maintenance and like maybe you spend

477
00:16:14,220 --> 00:16:18,180
some time building some cool

478
00:16:15,120 --> 00:16:20,880
visualizations and then we would sort of

479
00:16:18,180 --> 00:16:23,339
move on to other things

480
00:16:20,880 --> 00:16:24,899
and uh what actually happened was we

481
00:16:23,339 --> 00:16:27,180
launched the database and then we got

482
00:16:24,899 --> 00:16:29,220
this an overwhelming and unexpected

483
00:16:27,180 --> 00:16:32,160
response from the clinical Community

484
00:16:29,220 --> 00:16:34,139
because they are in the business of

485
00:16:32,160 --> 00:16:35,639
trying to figure out what variants what

486
00:16:34,139 --> 00:16:37,920
genetic variants that are observed in

487
00:16:35,639 --> 00:16:40,320
patients do and we had a whole bunch of

488
00:16:37,920 --> 00:16:41,639
experimental data about all of these

489
00:16:40,320 --> 00:16:43,139
different changes that happen in

490
00:16:41,639 --> 00:16:44,399
proteins many of which are clinically

491
00:16:43,139 --> 00:16:46,199
relevant and they're like well we want

492
00:16:44,399 --> 00:16:48,540
to use that to interpret what's

493
00:16:46,199 --> 00:16:49,860
happening in patients and I'm not going

494
00:16:48,540 --> 00:16:52,139
to tell you anything else about how

495
00:16:49,860 --> 00:16:54,420
clinical Genesis do the data use the

496
00:16:52,139 --> 00:16:55,920
data or how clinical genetics works but

497
00:16:54,420 --> 00:16:58,440
if you're interested please go check out

498
00:16:55,920 --> 00:17:01,079
David Warren's talk which I think is

499
00:16:58,440 --> 00:17:02,040
tomorrow at 11 20 a.m in Hall B and

500
00:17:01,079 --> 00:17:03,600
you'll learn more about the

501
00:17:02,040 --> 00:17:07,140
bioinformatics of variant classification

502
00:17:03,600 --> 00:17:08,160
and see some more Django stuff

503
00:17:07,140 --> 00:17:09,900
um

504
00:17:08,160 --> 00:17:11,280
so we also didn't get as much of a

505
00:17:09,900 --> 00:17:12,240
response from the researchers as we

506
00:17:11,280 --> 00:17:14,400
thought that we were going to and I

507
00:17:12,240 --> 00:17:16,079
think a lot of you saw that coming so we

508
00:17:14,400 --> 00:17:18,780
ended up having to take on some major

509
00:17:16,079 --> 00:17:21,179
curation curation tasks to get all the

510
00:17:18,780 --> 00:17:23,160
old papers in in particular and people

511
00:17:21,179 --> 00:17:25,799
are much more keen to deposit their new

512
00:17:23,160 --> 00:17:27,179
stuff rather than their old stuff and so

513
00:17:25,799 --> 00:17:29,040
this meant that we had to keep working

514
00:17:27,179 --> 00:17:30,900
on the project and invest a lot of time

515
00:17:29,040 --> 00:17:32,460
in developing new features and things

516
00:17:30,900 --> 00:17:34,380
like that

517
00:17:32,460 --> 00:17:36,740
and so all of these things that I told

518
00:17:34,380 --> 00:17:39,660
you before were still true

519
00:17:36,740 --> 00:17:41,460
but what became clear from working with

520
00:17:39,660 --> 00:17:43,980
uh starting to work with the clinical

521
00:17:41,460 --> 00:17:46,799
folks was that we needed really really

522
00:17:43,980 --> 00:17:48,720
good apis because we needed to make sure

523
00:17:46,799 --> 00:17:50,520
that once we had all this data in the

524
00:17:48,720 --> 00:17:52,799
central repository people did not want

525
00:17:50,520 --> 00:17:55,140
to come to our website to get it they

526
00:17:52,799 --> 00:17:57,720
wanted us to push it out to other

527
00:17:55,140 --> 00:18:00,299
biological biological data resources

528
00:17:57,720 --> 00:18:02,820
other front ends clinical Information

529
00:18:00,299 --> 00:18:05,600
Systems other places they wanted to be

530
00:18:02,820 --> 00:18:08,340
able to get it wholesale and this meant

531
00:18:05,600 --> 00:18:09,660
that we needed to have better API

532
00:18:08,340 --> 00:18:11,240
support than what we kept from Django

533
00:18:09,660 --> 00:18:13,620
out of the box because we needed to

534
00:18:11,240 --> 00:18:16,380
customize the data models combine

535
00:18:13,620 --> 00:18:17,520
internal data models in various ways and

536
00:18:16,380 --> 00:18:19,620
we also needed to do things like

537
00:18:17,520 --> 00:18:21,179
maintain different API endpoints for

538
00:18:19,620 --> 00:18:24,000
various collaborators based on their

539
00:18:21,179 --> 00:18:25,799
specific needs

540
00:18:24,000 --> 00:18:27,360
and meanwhile there was sort of trouble

541
00:18:25,799 --> 00:18:28,740
brewing with the project that those of

542
00:18:27,360 --> 00:18:30,299
you who are thinking about dates and

543
00:18:28,740 --> 00:18:32,039
have been around I've dreamed Django for

544
00:18:30,299 --> 00:18:34,200
a long time probably saw coming which

545
00:18:32,039 --> 00:18:37,020
was that uh you know a few months after

546
00:18:34,200 --> 00:18:39,000
we published the paper uh Django 1.11

547
00:18:37,020 --> 00:18:41,460
was no longer getting any security

548
00:18:39,000 --> 00:18:43,200
updates at all and then a little bit

549
00:18:41,460 --> 00:18:45,179
after that you know we ran out of money

550
00:18:43,200 --> 00:18:47,100
and the developer who had done a lot of

551
00:18:45,179 --> 00:18:48,299
really great stuff had also learned

552
00:18:47,100 --> 00:18:49,559
everything that he could learn on this

553
00:18:48,299 --> 00:18:50,880
project and got a really great

554
00:18:49,559 --> 00:18:53,220
opportunity decided you didn't want to

555
00:18:50,880 --> 00:18:55,080
do a PhD that he wanted to be a research

556
00:18:53,220 --> 00:18:56,940
software engineer and he got a great

557
00:18:55,080 --> 00:18:58,799
opportunity to go join a team and do

558
00:18:56,940 --> 00:19:01,679
some really cool stuff and so he left

559
00:18:58,799 --> 00:19:04,380
the project so now it's just me

560
00:19:01,679 --> 00:19:06,539
and Django version 1.11 and a live

561
00:19:04,380 --> 00:19:07,919
website that's running in production the

562
00:19:06,539 --> 00:19:09,720
guy who knew how everything worked is

563
00:19:07,919 --> 00:19:11,100
gone we could hire him back as a

564
00:19:09,720 --> 00:19:13,679
consultant a little bit but that's a

565
00:19:11,100 --> 00:19:16,080
very finite resource the GitHub security

566
00:19:13,679 --> 00:19:17,940
notifications kept getting spicier

567
00:19:16,080 --> 00:19:20,039
and we were trying to figure out what to

568
00:19:17,940 --> 00:19:21,240
do and so obviously this is not going to

569
00:19:20,039 --> 00:19:24,960
work so we need to do something that's

570
00:19:21,240 --> 00:19:27,240
not just maintaining the old Django site

571
00:19:24,960 --> 00:19:28,980
so we could migrate to Django 2.2

572
00:19:27,240 --> 00:19:30,840
long-term support but then we would just

573
00:19:28,980 --> 00:19:31,740
have this problem again later on in not

574
00:19:30,840 --> 00:19:33,900
too long

575
00:19:31,740 --> 00:19:36,120
we could move to Django 3 but all the

576
00:19:33,900 --> 00:19:37,559
information we could find said this

577
00:19:36,120 --> 00:19:39,059
would be a lot of work and it was kind

578
00:19:37,559 --> 00:19:40,679
of new and scary and it also wasn't

579
00:19:39,059 --> 00:19:42,840
clear that we would get the kind of API

580
00:19:40,679 --> 00:19:45,059
support that we needed which again was a

581
00:19:42,840 --> 00:19:47,700
requirement that we didn't know that we

582
00:19:45,059 --> 00:19:50,760
had when we started the project but now

583
00:19:47,700 --> 00:19:53,280
was kind of our major requirement or we

584
00:19:50,760 --> 00:19:54,900
could try something new whatever that is

585
00:19:53,280 --> 00:19:56,520
now fortunately things worked out

586
00:19:54,900 --> 00:19:58,860
because while we were deciding what to

587
00:19:56,520 --> 00:20:00,840
do we ended up and doing a lot of this

588
00:19:58,860 --> 00:20:02,640
additional data curation work we ended

589
00:20:00,840 --> 00:20:05,039
up being successful in some major grant

590
00:20:02,640 --> 00:20:06,179
funding rounds in the United States this

591
00:20:05,039 --> 00:20:07,500
allowed me to hire a new Junior

592
00:20:06,179 --> 00:20:08,820
developer in Melbourne so I wasn't

593
00:20:07,500 --> 00:20:11,039
trying to do it all on my own anymore

594
00:20:08,820 --> 00:20:13,020
and also the host institution that was

595
00:20:11,039 --> 00:20:15,059
running the server finished their

596
00:20:13,020 --> 00:20:17,940
upgrades to Centos 7 just a little bit

597
00:20:15,059 --> 00:20:19,799
after the end of Centos six and we

598
00:20:17,940 --> 00:20:22,140
managed to get permission to run Docker

599
00:20:19,799 --> 00:20:23,640
compose which really simplified our

600
00:20:22,140 --> 00:20:25,620
devops challenges and made it much

601
00:20:23,640 --> 00:20:28,440
easier to kind of get all of these

602
00:20:25,620 --> 00:20:30,299
things to play well together

603
00:20:28,440 --> 00:20:32,460
[Applause]

604
00:20:30,299 --> 00:20:34,140
uh and then we got some huge help when a

605
00:20:32,460 --> 00:20:35,460
senior developer in Seattle who was

606
00:20:34,140 --> 00:20:37,740
working with one of the collaborating

607
00:20:35,460 --> 00:20:40,260
labs was able to contribute some time to

608
00:20:37,740 --> 00:20:41,640
the project and help us figure out a

609
00:20:40,260 --> 00:20:44,580
path forward

610
00:20:41,640 --> 00:20:46,500
so now that we had this experience and

611
00:20:44,580 --> 00:20:48,600
we had some senior people who had their

612
00:20:46,500 --> 00:20:50,580
own preferences we were able to take a

613
00:20:48,600 --> 00:20:53,580
look at this again and what we decided

614
00:20:50,580 --> 00:20:54,900
to do was move from Django 1.11 to fast

615
00:20:53,580 --> 00:20:56,700
API

616
00:20:54,900 --> 00:20:58,620
and even though fast API is obviously

617
00:20:56,700 --> 00:21:00,720
very different from Django the fact that

618
00:20:58,620 --> 00:21:02,220
all the application specific code that

619
00:21:00,720 --> 00:21:04,020
we wrote for data set validation

620
00:21:02,220 --> 00:21:06,179
handling genomic information all of

621
00:21:04,020 --> 00:21:08,880
these other tasks was written in Python

622
00:21:06,179 --> 00:21:10,860
and using pandas and other libraries it

623
00:21:08,880 --> 00:21:12,780
meant that most of what we had done we

624
00:21:10,860 --> 00:21:14,700
could just bring straight over and we

625
00:21:12,780 --> 00:21:16,679
were also even able to use the same

626
00:21:14,700 --> 00:21:18,660
postgres database that we had we didn't

627
00:21:16,679 --> 00:21:20,580
have to really I mean obviously we had

628
00:21:18,660 --> 00:21:22,500
to do some migrations but but that

629
00:21:20,580 --> 00:21:26,160
really pretty much got to stay the same

630
00:21:22,500 --> 00:21:28,500
we just shifted what was on top of it

631
00:21:26,160 --> 00:21:30,059
so now as of uh this sort of the new

632
00:21:28,500 --> 00:21:31,620
version here's what the Technology stock

633
00:21:30,059 --> 00:21:33,059
looks like we've gone from you know

634
00:21:31,620 --> 00:21:34,740
having to figure out how to get

635
00:21:33,059 --> 00:21:37,140
everything working under a specific

636
00:21:34,740 --> 00:21:39,419
version of Centos to doing things in

637
00:21:37,140 --> 00:21:41,340
Docker compose which is way nicer both

638
00:21:39,419 --> 00:21:43,320
for you know deploying it and also for

639
00:21:41,340 --> 00:21:44,580
development it's much easier to set up a

640
00:21:43,320 --> 00:21:46,559
development in a local development

641
00:21:44,580 --> 00:21:48,480
environment now we upgraded the version

642
00:21:46,559 --> 00:21:50,760
of postgresql that we were doing which

643
00:21:48,480 --> 00:21:52,320
is great because now our awesome Json

644
00:21:50,760 --> 00:21:54,179
fields are now even more awesome and

645
00:21:52,320 --> 00:21:57,000
more performant we got to move from

646
00:21:54,179 --> 00:22:00,419
Python 3.4 to 3.9 which made me very

647
00:21:57,000 --> 00:22:02,880
happy and then we also switched from

648
00:22:00,419 --> 00:22:07,460
using Django to using a combination of

649
00:22:02,880 --> 00:22:07,460
fast API and view

650
00:22:07,919 --> 00:22:12,059
and so this really took a lot of work

651
00:22:09,840 --> 00:22:13,980
but we're now basically back to where

652
00:22:12,059 --> 00:22:15,840
the Django site was but we rebuilt it in

653
00:22:13,980 --> 00:22:18,659
the new framework and our hope is that

654
00:22:15,840 --> 00:22:21,720
we've sort of built kind of a new floor

655
00:22:18,659 --> 00:22:23,220
um where where where we where we were

656
00:22:21,720 --> 00:22:25,140
before and now we can start adding some

657
00:22:23,220 --> 00:22:27,299
new features so time will tell and this

658
00:22:25,140 --> 00:22:28,799
took about a year so we started in you

659
00:22:27,299 --> 00:22:31,140
know the first commit in the new version

660
00:22:28,799 --> 00:22:33,240
was in April of 2022 and then in April

661
00:22:31,140 --> 00:22:36,659
of this year we switched over from the

662
00:22:33,240 --> 00:22:37,980
Django version to the new version

663
00:22:36,659 --> 00:22:40,080
um

664
00:22:37,980 --> 00:22:42,059
and so I think the moral of the story is

665
00:22:40,080 --> 00:22:43,620
this and which has been said by by a lot

666
00:22:42,059 --> 00:22:45,059
of other folks in different contexts

667
00:22:43,620 --> 00:22:46,320
which is that it's impossible to know

668
00:22:45,059 --> 00:22:48,179
what to build until you've tried to

669
00:22:46,320 --> 00:22:49,980
build something or build something else

670
00:22:48,179 --> 00:22:51,480
first and it's really held true for all

671
00:22:49,980 --> 00:22:53,520
of the software projects that I've done

672
00:22:51,480 --> 00:22:54,299
as a researcher

673
00:22:53,520 --> 00:22:57,360
um

674
00:22:54,299 --> 00:22:59,400
and I can confidently say that uh

675
00:22:57,360 --> 00:23:01,380
you know having a framework like Django

676
00:22:59,400 --> 00:23:03,120
available that we could pick up and use

677
00:23:01,380 --> 00:23:04,980
to get something built was absolutely

678
00:23:03,120 --> 00:23:06,600
essential for us to succeed in this

679
00:23:04,980 --> 00:23:08,820
project and while we were getting

680
00:23:06,600 --> 00:23:10,679
started we absolutely did not have the

681
00:23:08,820 --> 00:23:12,179
capacity or the funding to take on a

682
00:23:10,679 --> 00:23:14,760
project that required expertise into

683
00:23:12,179 --> 00:23:17,940
programming languages

684
00:23:14,760 --> 00:23:19,679
um we didn't have anywhere near a clear

685
00:23:17,940 --> 00:23:21,840
enough idea of what we needed to build

686
00:23:19,679 --> 00:23:24,780
in order to take advantage of a sort of

687
00:23:21,840 --> 00:23:27,240
a more flexible and open-ended uh kind

688
00:23:24,780 --> 00:23:29,400
of development uh environment like uh

689
00:23:27,240 --> 00:23:30,840
like we're getting with fast API and

690
00:23:29,400 --> 00:23:32,880
view

691
00:23:30,840 --> 00:23:34,620
um and most importantly I think no

692
00:23:32,880 --> 00:23:36,360
amount of us thinking hard about the

693
00:23:34,620 --> 00:23:38,940
problem would have gotten us to the

694
00:23:36,360 --> 00:23:40,799
point that we needed to get to because

695
00:23:38,940 --> 00:23:43,200
we needed the engagement from external

696
00:23:40,799 --> 00:23:44,940
parties uh with different needs and

697
00:23:43,200 --> 00:23:48,140
different backgrounds and we could only

698
00:23:44,940 --> 00:23:50,880
get that by deciding to build something

699
00:23:48,140 --> 00:23:53,700
building it and putting it out into the

700
00:23:50,880 --> 00:23:56,520
world and then seeing what happened

701
00:23:53,700 --> 00:23:58,260
um and so you know we would not have had

702
00:23:56,520 --> 00:24:01,740
success if we had started with the

703
00:23:58,260 --> 00:24:03,480
framework that we're using now and so we

704
00:24:01,740 --> 00:24:05,159
think that this project despite the fact

705
00:24:03,480 --> 00:24:06,659
that we're not currently using Django is

706
00:24:05,159 --> 00:24:08,059
a Django success story and I hope that

707
00:24:06,659 --> 00:24:11,159
you'll agree that that's the case

708
00:24:08,059 --> 00:24:15,299
because uh we never would have gotten to

709
00:24:11,159 --> 00:24:18,299
where we are without Django and I would

710
00:24:15,299 --> 00:24:20,159
really like to thank Django Khan and the

711
00:24:18,299 --> 00:24:22,679
and this audience for giving me the

712
00:24:20,159 --> 00:24:25,020
opportunity to share it with you

713
00:24:22,679 --> 00:24:26,700
um and with that here's some some links

714
00:24:25,020 --> 00:24:28,559
and references if you want the slides

715
00:24:26,700 --> 00:24:29,940
I'll figure out how to give you the

716
00:24:28,559 --> 00:24:32,760
slides

717
00:24:29,940 --> 00:24:34,440
um there's some some links to the to the

718
00:24:32,760 --> 00:24:36,840
papers that were linked in there and

719
00:24:34,440 --> 00:24:38,159
then also the website itself and and our

720
00:24:36,840 --> 00:24:38,820
GitHub

721
00:24:38,159 --> 00:24:41,400
um

722
00:24:38,820 --> 00:24:44,280
and also add that if if anybody wants

723
00:24:41,400 --> 00:24:46,080
papers and you you can just email the

724
00:24:44,280 --> 00:24:47,700
author that's listed and we can send it

725
00:24:46,080 --> 00:24:50,700
to you for free and we love to do that

726
00:24:47,700 --> 00:24:53,100
so if you want biological literature

727
00:24:50,700 --> 00:24:54,780
just just ask the scientists and with

728
00:24:53,100 --> 00:24:57,620
that I would be very happy to take any

729
00:24:54,780 --> 00:24:57,620
questions if there's time

730
00:25:04,020 --> 00:25:08,760
so much for that Alan we have the

731
00:25:06,840 --> 00:25:10,980
Discord where we will take questions I'm

732
00:25:08,760 --> 00:25:13,380
happy to read those out you can do it on

733
00:25:10,980 --> 00:25:15,900
mobile and stuff but it's been a couple

734
00:25:13,380 --> 00:25:18,840
of years so now I need to remember how

735
00:25:15,900 --> 00:25:22,260
fast I can run to get Russell the first

736
00:25:18,840 --> 00:25:24,659
question because of course it is

737
00:25:22,260 --> 00:25:26,700
hey we're back

738
00:25:24,659 --> 00:25:28,380
is it reporting you

739
00:25:26,700 --> 00:25:30,539
um thanks for the talk this is fantastic

740
00:25:28,380 --> 00:25:31,919
and I'd love to hear success stories of

741
00:25:30,539 --> 00:25:34,740
Django anytime

742
00:25:31,919 --> 00:25:39,179
um could you speak to what features of

743
00:25:34,740 --> 00:25:41,400
fast API led to that transition over and

744
00:25:39,179 --> 00:25:44,460
if they are things that Django as a

745
00:25:41,400 --> 00:25:46,200
project as a community could adopt to

746
00:25:44,460 --> 00:25:49,260
like not that you made the wrong choice

747
00:25:46,200 --> 00:25:51,000
but but to prevent other people making

748
00:25:49,260 --> 00:25:52,620
that choice in the future to make to

749
00:25:51,000 --> 00:25:53,820
make Django more appealing to the to the

750
00:25:52,620 --> 00:25:55,919
demographic that we're not currently

751
00:25:53,820 --> 00:25:57,059
satisfying sure

752
00:25:55,919 --> 00:26:00,360
um

753
00:25:57,059 --> 00:26:03,299
so I think that one of the in our

754
00:26:00,360 --> 00:26:05,940
experience in in Django one was that the

755
00:26:03,299 --> 00:26:07,679
the we got a rest API out of the box but

756
00:26:05,940 --> 00:26:10,860
that was essentially exposing the

757
00:26:07,679 --> 00:26:13,440
internal data models and so

758
00:26:10,860 --> 00:26:15,120
changing the API Behavior then would

759
00:26:13,440 --> 00:26:17,279
either require us to write a whole bunch

760
00:26:15,120 --> 00:26:19,320
of code to kind of build a custom API

761
00:26:17,279 --> 00:26:23,820
which we didn't really want to do

762
00:26:19,320 --> 00:26:26,100
or it would require us to be able to

763
00:26:23,820 --> 00:26:28,500
um go back and refactor those basic

764
00:26:26,100 --> 00:26:31,039
models which I think would have been

765
00:26:28,500 --> 00:26:35,220
not also not feasible

766
00:26:31,039 --> 00:26:37,020
so by using fast API and pedantic and

767
00:26:35,220 --> 00:26:39,659
the sort of you know models versus view

768
00:26:37,020 --> 00:26:41,700
models we can build view models which

769
00:26:39,659 --> 00:26:43,020
are relatively easy to consume we can

770
00:26:41,700 --> 00:26:46,580
have a lot of different flavors of view

771
00:26:43,020 --> 00:26:50,600
models and then build the API using that

772
00:26:46,580 --> 00:26:52,980
and that was a really uh kind of the the

773
00:26:50,600 --> 00:26:54,720
main advantage technical advantage that

774
00:26:52,980 --> 00:26:56,520
we were getting was that that ability to

775
00:26:54,720 --> 00:26:59,279
put sort of whatever it is that we

776
00:26:56,520 --> 00:27:01,799
needed in those view models and have it

777
00:26:59,279 --> 00:27:03,659
pretty much just work and even build in

778
00:27:01,799 --> 00:27:05,940
some behavior and other stuff in those

779
00:27:03,659 --> 00:27:07,200
it's also a cool nice I mean one of the

780
00:27:05,940 --> 00:27:09,840
things that I didn't mention is we built

781
00:27:07,200 --> 00:27:11,760
like sort of an SDK kind of thing so

782
00:27:09,840 --> 00:27:13,440
that people can just write python code

783
00:27:11,760 --> 00:27:15,720
and be able to upload and download data

784
00:27:13,440 --> 00:27:17,460
sets so people who are you know you're

785
00:27:15,720 --> 00:27:19,860
sort of like I did a python boot camp

786
00:27:17,460 --> 00:27:21,360
and you know data scientists like wet

787
00:27:19,860 --> 00:27:22,380
lab people

788
00:27:21,360 --> 00:27:24,419
um

789
00:27:22,380 --> 00:27:25,740
and and so it's nice being able we

790
00:27:24,419 --> 00:27:28,260
connect we actually import The View

791
00:27:25,740 --> 00:27:29,700
models into that and do to do vocal

792
00:27:28,260 --> 00:27:31,440
validation and so there's some kind of

793
00:27:29,700 --> 00:27:33,720
other nice features there so that's

794
00:27:31,440 --> 00:27:37,159
that's I think the main thing

795
00:27:33,720 --> 00:27:37,159
you have time for one more question

796
00:27:43,500 --> 00:27:48,179
um uh the is the experimental data

797
00:27:46,440 --> 00:27:49,860
consistent like is it like a

798
00:27:48,179 --> 00:27:53,820
well-established file format for the

799
00:27:49,860 --> 00:27:54,960
data Etc not at all not even close uh

800
00:27:53,820 --> 00:27:58,140
one of the cool things about this

801
00:27:54,960 --> 00:28:00,360
technology is that it's a very

802
00:27:58,140 --> 00:28:02,159
versatile technology and anyone who's

803
00:28:00,360 --> 00:28:03,179
interested in studying proteins whether

804
00:28:02,159 --> 00:28:04,860
that's someone who's interested in

805
00:28:03,179 --> 00:28:06,299
approaching engineering or someone who's

806
00:28:04,860 --> 00:28:07,440
interested in clinical genetics or

807
00:28:06,299 --> 00:28:09,840
someone who's interested in evolution

808
00:28:07,440 --> 00:28:11,100
they're all converging on using this

809
00:28:09,840 --> 00:28:12,840
technology and then they're publishing

810
00:28:11,100 --> 00:28:14,760
their papers and so they're all coming

811
00:28:12,840 --> 00:28:17,820
from different fields with different

812
00:28:14,760 --> 00:28:19,559
backgrounds and different sort of data

813
00:28:17,820 --> 00:28:22,159
conventions and data sharing conventions

814
00:28:19,559 --> 00:28:24,840
and so it is a hot mess

815
00:28:22,159 --> 00:28:27,000
and that's why it's even more important

816
00:28:24,840 --> 00:28:28,980
to have a database that can put all this

817
00:28:27,000 --> 00:28:31,440
stuff in one place so that then the

818
00:28:28,980 --> 00:28:33,600
computer scientists can go there and not

819
00:28:31,440 --> 00:28:36,779
have to learn all this different biology

820
00:28:33,600 --> 00:28:40,320
to you know get it get it what they want

821
00:28:36,779 --> 00:28:43,820
so and with that we're out of time we

822
00:28:40,320 --> 00:28:43,820
would like to thank Alan again

823
00:28:45,490 --> 00:28:48,869
[Applause]