From fbc7ec1f342767052e23d6430ea94f0ae7c8d27f Mon Sep 17 00:00:00 2001 From: Christoph Helma Date: Fri, 22 Jan 2016 16:55:39 +0100 Subject: rendered workshop presentation --- presentation/enm-workshop.html | 445 +++++++++++++++++++++++++++++++++++++++++ 1 file changed, 445 insertions(+) create mode 100644 presentation/enm-workshop.html (limited to 'presentation/enm-workshop.html') diff --git a/presentation/enm-workshop.html b/presentation/enm-workshop.html new file mode 100644 index 0000000..f0c6781 --- /dev/null +++ b/presentation/enm-workshop.html @@ -0,0 +1,445 @@ + + + + + + + +Read across toxicity predictions with nano-lazar + + + + + + + + + + + + + + +
+
+
+ + +
+
+
+

Read across toxicity predictions with nano-lazar

+ +

Christoph Helma

+

in silico toxicology gmbh

+http://www.enanomapper.net/sites/all/themes/theme807/logo.png + +
+
+

Requirements

+
    +
  • Nanoparticle characterisation
    • +
  • Toxicity measurements
    • +
+
+
+

eNanoMapper data import

+
    +
  • Nanoparticles imported: 464
    • +
  • Nanoparticles with particle characterisation: 394
    • +
  • Nanoparticles with toxicity data: 167
    • +
  • Nanoparticles with toxicity data and particle characterisation: 160
    • +
+
+
+

eNanoMapper toxicity endpoints

+
    +
  • Toxicity endpoints: 41
    • +
  • Toxicity endpoints with more than one measurement value: 22
    • +
  • Toxicity endpoints with more than 10 measurements: 2
    • +
+
+
+

Selected data

+

Protein corona dataset Au particles (106 particles) +Toxicity endpoint:

+
+
+

Read across procedure

+
    +
  • Identify relevant fragments (significant correlation with toxicity) +TODO list of fragments, number
    • +
  • Calculate similarities (weighted cosine similarity, correlation coefficients = weights)
    • +
  • Identify neighbors (particles with more than 0.95 similarity)
    • +
  • Calculate prediction (weighted average from neighbors, similarities = weights)
    • +
+
+
+

Future development

+
    +
  • Validation of predictions
    • +
  • Applicability domain/reliability of predictions
    • +
  • Accuracy improvements: +- additional data +- feature selection +- similarity calculation +- predictions (local regression models)
    • +
  • Usability improvements: +- additional data (extension of applicability domain, additional endpoints and chemistries) +- inclusion of ontologies +- descriptor calculation directly from core and coating chemistries
    • +
+
+
+

Webinterface

+

https://nano-lazar.in-silico.ch/predict

+

Your recommendations?

+
+ +
+ + -- cgit v1.2.3