O PTIMAL NANO - DESCRIPTORS AS TRANSLATORS OF ECLECTIC DATA INTO PREDICTION OF THE CELL MEMBRANE DAMAGE BY MEANS OF NANO METAL - OXIDES A LLA P. T OROPOVA & A NDREY A. T OROPOV & E MILIO B ENFENATI & R AFI K ORENSTEIN & D ANUTA L ESZCZYNSKA & J ERZY L ESZCZYNSKI Environ Sci Pollut Res (2015) 22:745–757 Acknowledgments: The authors are grateful to the EU FP7 project PreNanoTox (contract ).

S YSTEMATIZATION OF KNOWLEDGE ON NANOMATERIALS HAS BECOME A NECESSITY WITH THE FAST GROWTH OF APPLICATIONS OF THESE SPECIES. B UILDING UP PREDICTIVE MODELS THAT DESCRIBE PROPERTIES ( BOTH BENEFICIAL AND HAZARDOUS ) OF NANOMATERIALS IS VITAL FOR COMPUTATIONAL SCIENCES. C LASSIC QUANTITATIVE STRUCTURE – PROPERTY / ACTIVITY RELATIONSHIPS (QSPR/QSAR) ARE NOT SUITABLE FOR INVESTIGATING NANOMATERIALS BECAUSE OF THE COMPLEXITY OF THEIR MOLECULAR ARCHITECTURE. H OWEVER, SOME CHARACTERISTICS SUCH AS SIZE, CONCENTRATION, AND EXPOSURE TIME CAN INFLUENCE ENDPOINTS ( BENEFICIAL OR HAZARDOUS ) RELATED TO NANOPARTICLES AND THEY CAN THEREFORE BE INVOLVED IN BUILDING A MODEL. A PPLICATION OF THE OPTIMAL DESCRIPTORS CALCULATED WITH THE SO - CALLED CORRELATION WEIGHTS OF VARIOUS CONCENTRATIONS AND DIFFERENT EXPOSURE TIMES ARE SUGGESTED IN ORDER TO BUILD UP A PREDICTIVE MODEL FOR CELL MEMBRANE DAMAGE CAUSED BY A SERIES OF NANO METAL - OXIDES. T HE NUMERICAL DATA ON CORRELATION WEIGHTS ARE CALCULATED BY THE M ONTE C ARLO METHOD. T HE OBTAINED RESULTS ARE IN GOOD AGREEMENT WITH THE EXPERIMENTAL DATA.

T HE EXPERIMENTAL DATA ON CELLMEMBRANE DAMAGEMEASURED BY PROPIDIUM IODIDE (PI) UPTAKE ARE TAKEN FROM THE LITERATURE, WHICH ALSO REPORTS 24 NANO METAL - OXIDES (Z R O2, Z N O, Y B 2O3, Y2O3, WO3, T I O2, S N O2, S I O2, S B 2O3, N I O, N I 2O3, M N O3, L A 2O3, I N 2O3, H F O2, G D 2O3, F E 3O4, F E 2O3, C U O, C R 2O3, C O O, C O 3O4, C E O2, A L 2O3). T HE NUMERICAL DATA ON THIS ENDPOINT RELATED TO FOUR DOSES (50, 100, 150, AND 200 ΜG / M L) AND SEVEN EXPOSURE TIME ( FROM 1 TO 7 H ) FOR ALL 24 NANO METAL - OXIDES ARE EXAMINED. I N FACT, THE PERCENTAGE OF CELLS WHICH HAVE MEMBRANE DAMAGE IS THE MEASURE OF IMPACT OF NANO - OXIDES ( FOR DEFINED DOSE AND EXPOSURE TIME ).

Q UASI -SMILES FOR REPRESENTATION OF IMPACT OF NANO -F E 3O4 UPON CELL MEMBRANE, WHERE DOSE IS 200 ΜG / M L (“A”) AND EXPOSURE TIME IS 5 H (“5”)

T HE LIST OF C K WHICH ARE USED FOR REPRESENTATION OF NANO METALOXIDES AND CONDITIONS OF THEIR ACTING. F OR EXAMPLE, THE “Z N.O.A2” MEANS ( I ) NANO METAL - OXIDE Z N O, ( II ) DOSE IS 200 ΜG / M L, AND ( III ) EXPOSURE TIME IS 2 H

H AVING DATA ON OPTIMAL CORRELATION WEIGHTS, ONE CAN : ( I ) CALCULATE DCW(T,N EPOCH ) FOR ALL NANOMETAL - OXIDES ; ( II ) CALCULATE ( WITH DATA ON THE TRAINING SET ) A MODEL FOR CELL MEMBRANE DAMAGE (CMD) AND ( III ) CHECK UP THE PREDICTIVE POTENTIAL OF THE MODEL USING THE EXTERNAL VALIDATION SET. The T and Nepoch are parameters of the optimization: the T is threshold, i.e., the coefficient for classification of CAk into two categories: rare and not rare. The correlation weight for rare impact is set as zero, so this component is not involved in a model. The Nepoch is the number of epochs of the Monte Carlo optimization. The computational experiments with the CORAL software have shown that as a rule, for an endpoint preferable, T* and N*epoch can be extracted after analysis of a range of the T (e.g., from 1 to 10) and analysis of a range of Nepoch (e.g., from 1 to 100).

T HE SCHEME OF CALCULATION OF PREFERABLE VALUES T* AND N* EPOCH

… … … … …… … …..

T HE CELL MEMBRANE DAMAGE BY MEANS OF NANO METAL - OXIDES IS A COMPLEX FUNCTION OF MANY FACTORS BESIDES CHEMICAL COMPOSITION, DOSE, AND EXPOSURE TIME. H OWEVER, THE COMPARISON OF FIVE MODELS, WHICH ARE BUILT UP FOR DIFFERENT DISTRIBUTION INTO THE VISIBLE TRAINING SET AND INVISIBLE VALIDATION SET INDICATES THAT CHEMICAL COMPOSITION, DOSE, AND EXPOSURE TIME CAN BE A BASIS TO PREDICT THE ENDPOINT USING THE M ONTE C ARLO METHOD. Conclusions:

Appendix: Screen Shot Of Step 1

Screen Shot Of Step 2

Screen Shot Of Step 3

Screen Shot Of Step 4

Screen Shot Of Step 5 5