1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
|
=========================================
In silico methods for toxicity prediction
=========================================
:Author: Christoph Helma
:Affiliation: in silico toxicology gmbh
:Date: 2012-10-04
.. footer:: 2012-10-04 *in silico* toxicology gmbh
Outline
=======
.. class:: incremental
- In silico toxicology methods
- Lazar framework
- Products and services
In silico methods
=================
.. class:: incremental
- Systems biology/molecular modeling
- Expert systems
- Data driven techniques
Systems biology/molecular modeling
====================================
Model individual events (e.g. receptor interactions, (de)toxification) of the *adverse outcome pathway*
.. class:: incremental small green
- Mechanistic interpretation
.. class:: incremental small red
- Calculations/simulations may be very time consuming
- May require a lot of experimental data for parameterization
- Impossible to model mechanisms of complex toxicological endpoints
Examples: VirtualToxLab/Biograf
Expert systems
==============
Formalize expert knowledge about chemicals and toxicity mechanisms and create a software program
.. class:: incremental small green
- Mechanistic interpretation
.. class:: incremental small red
- Model creation very time consuming
- Many toxicity mechanisms are poorly understood or even unknown
- Error prone and hard to validate (strong tendency towards overfitting)
Examples: Derek/Lhasa
Data driven
===========
Use all existing data for a particular endpoint and apply machine learning/QSAR algorithms in order to create a prediction model
.. class:: incremental small green
- Comparably fast
- Applicable for every endpoint with sufficient experimental data
- Sound validation possible
.. class:: incremental small red
- Applicability domain/model quality depends on experimental data
- Mechanistic relevance has to be extracted from models/descriptors/predictions
Examples: Classical QSARs, Topkat, Multicase, lazar
Lazy-Structure-Activity Relationships (lazar)
=============================================
Automated read across predictions
.. class:: incremental small
- Find *similar* compounds (=neighbors) with measured activities
- Create a local (Q)SAR model with neighbors as training compounds
- Make a prediction with this model
.. class:: incremental
Lazar estimates the confidence (*applicability domain*) for each prediction
Chemical Similarity
===================
Can be based on
.. class:: incremental
- Chemical structures
- Chemical properties
- Biological properties
- ...
.. class:: incremental
Lazar uses *activity specific similarities*
Activity specific similarities
==============================
Consider only *relevant* (i.e. statistically significant) substructures, properties, ... for similarity calculations
Algorithms for finding relevant substructures (by A. Maunz):
- Backbone refinement classes (BBRC)
- Latent structure mining (LAST)
http://lazar.in-silico.ch
=========================
.. image:: lazar-input.png
:height: 13em
:align: center
http://lazar.in-silico.ch
=========================
.. image:: lazar-output1.png
:height: 13em
:align: center
http://lazar.in-silico.ch
=========================
.. image:: lazar-output2.png
:height: 13em
:align: center
http://lazar.in-silico.ch
=========================
.. image:: lazar-output3.png
:height: 13em
:align: center
Lazar limitations
=================
.. class:: incremental
- Model quality depends on data quality
- Applicability domain depends on learning instances
in silico toxicology gmbh
=========================
Open source software and algorithm development
.. class:: incremental
- Predictive toxicology and QSAR models
- Toxicological data mining
- Life science webservices and data warehouses
Why open source?
================
.. class:: incremental
- Clear and unambiguous documentation of implemented algorithms essential for scientific software (also required by many regulatory guidelines)
- Collaboration with partners, projects and external contributors
- Establishment of international standards
- Security of investment
EU Research projects (FP6/7)
============================
.. class:: small
:Sens-it-iv: Novel testing strategies for in vitro assessment of allergens
:Scarlet: Network on in silico methods for carcinogenicity and mutagenicity
:OpenTox: Open source framework for predictive toxicology
:ToxBank: Integrated data analysis and servicing of alternative testing methods in toxicology
:ModNanoTox: Modelling toxicity behaviour of engineered nanoparticles
Free products and services
==========================
:Lazar application: http://lazar.in-silico.ch
:OpenTox Webservices: http://webservices.in-silico.ch
:Source code: https://github.com/opentox
Issue tracker, documentation, ...
Commercial products and services
================================
.. class:: incremental
- Lazar "software as a service" (SaaS): secure access for confidential predictions, batch predictions, ...
- Virtual appliances with lazar software for in-house/desktop installation
- Installation services
- Phone and email support
Commercial products and services
================================
.. class:: incremental
- Virtual toxicity screening of compounds and libraries
- Development of prediction models for new endpoints
- Scientific programming, contract research and consulting
Contact
=======
:Web: http://www.in-silico.ch
:Email: helma@in-silico.ch
|
