Skip to content

Latest commit

 

History

History
119 lines (81 loc) · 3.51 KB

malaga-presentation.md

File metadata and controls

119 lines (81 loc) · 3.51 KB
Raw

% nano-lazar: A framework for nanoparticle read across risk assessment % Christoph Helma, Micha Rautenberg, Denis Gebele % in silico toxicology gmbh, Basel, Switzerland


nano-lazar framework

Framework for reproducible read-across (k-nearest-neighbor) predictions

For a given query substance (nanoparticle)

  • Find similar substances (neighbors)
  • Create local model with neighbor activities
  • Apply local model to predict the query substance

nano-lazar algorithms

Free choice of algorithms and parameters for

  • Descriptors (measured, calculated)
  • Feature selection
  • Similarity calculation
  • Local QSAR models

Reasonable default algorithms and parameters

nano-lazar experiments

Endpoint: ~ Net cell association (121 Gold and Silver particles)

Descriptors: ~ Calculated fingerprints (new development), measured physchem properties, protein interactions

Feature selection: ~ Correlation filter (measured properties)

Similarity: ~ Weighted cosine (measured properties), Tanimoto/Jaccard (fingerprints)

Local regression: ~ Weighted average, partial least squares (PLS), random forests (RF)

  • Five independent 10-fold crossvalidations
  • No fixed random seed for training/test set splits, to avoid overfitting and to demonstrate the variability of validation results due to random training/test splits.
  • Separate feature selection for each training dataset to avoid overfitting

nano-lazar validation results

m

Gold and silver particles included!

nano-lazar prediction/measurement correlation

![](figures/Proteomics-rf-0.jpeg){#fig:prot0 width=18%} ![](figures/Proteomics-rf-1.jpeg){#fig:prot1 width=18%} ![](figures/Proteomics-rf-2.jpeg){#fig:prot2 width=18%} ![](figures/Proteomics-rf-3.jpeg){#fig:prot3 width=18%} ![](figures/Proteomics-rf-4.jpeg){#fig:prot4 width=18%}

Correlation of predicted vs. measured values for five independent crossvalidations with Proteomics descriptors and local random forest models

nano-lazar validation summary

  • Best results: Protein interaction descriptors and random forest models
  • Most results show no statistically significant difference
  • Calculated fingerprints perform surprisingly well (RMSE comparable to best results) and do not require measured nanoparticle properties
  • Local random forest models with nanoparticle fingerprints can be useful for screening purposes without physical measurements

nano-lazar GUI

![](images/nano-lazar fingerprint input.ch_predict.png){#id .class height=50%}

nano-lazar GUI

![](images/nano-lazar fingerprint result.ch_predict.png){#id .class height=50%}

Reproducible research

Manuscript submitted to Frontiers in Pharmacology

  • Open source code for lazar and publication
  • Manuscript with direct source code links
  • Docker image with
    • lazar libraries
    • Training data
    • Validation experiments
    • Software dependencies
    • Build system to recreate validations
    • Update facilities for code and data

Links

Nano-lazar GUI : https://nano-lazar.in-silico.ch

Lazar (source code) : https://github.com/opentox/lazar

Publication (source code) : https://github.com/opentox/nano-lazar-paper

Docker image : https://hub.docker.com/r/insilicotox/nano-lazar-paper/

Twitter ~ https://twitter.com/insilicotox

Email ~ [email protected]