School of Chemistry, UNSW

Chemometrics And Metrology Group (CMG) - Group Members

Diako Ebrahimi, Postdoc - 2007

Born in 1977, Ph.D. from UNSW (2007), married with one child.

Address: Room 123, Dalton Building, School of Chemistry, UNSW

Tel: +61 2 9385 6621

Fax: +61 2 9385 6141

E-mail: Diako@unsw.edu.au

Cv: PDF

Dr Ebrahimi is working as a research associate in the Chemometrics group on various multi-way analysis projects including:

Source identification of oil spills

GC-MS data including those of petroleum oils are naturally three-way (GC x MS x SAMPLE) and therefore are perfect candidates for analysis by multi-way models such as PARAFAC2 which can also accommodate shifts in the chromatograms. We are interested in identification of source of oil spills, in particular lighter oils such as diesels, and exploring less known degradation mechanisms such as photo-oxidation using multi-way analysis.

C1-Phenanthrene (C1-Ph) extracted ion chromatogram of a crude oil and a suspect source oil. Five structural isomers of C1-Ph are shown by numbers 1-, 2-, 3-, 4- and 9- referring to the position of methyl substitute on the phenanthrene molecule. In this example all the isomers are present in both petroleum oils however because the pattern of the EICs (relative intensity of isomer peaks) are different, the suspect source is a "non-match" hit.

Schematic presentation of a three-way GC-MS data which can be modelled by PARAFAC2 but not by PARAFAC. a) GC-MS data with chromatographic shifts from sample to sample, b) GC-MS data with different elution time dimension in each m/z.

See our recent publication for more details.

Analysis of complex interactions

Estimation of large number of parameters to model higher order interaction terms limits the interpretability and therefore applicability of classic ANOVA models. Multiplicative models have been proposed to tackle this problem in data generated mainly by interactions. Generalized Multiplicative Analysis of Variance (GEMANOVA) method is a very useful method that can be applied to analyse experimentally designed data (i.e. multi-way data) from complex systems such as those from biology. In our group we investigate various aspects of GEMANOVA and similar models and their applications.

Presentation of a GEMANOVA model with four terms using a constrained three-way PARAFAC model with four trilinear components. Each trilinear component in PARAFAC corresponds to a term in GEMANOVA. GEMANOVA interaction terms with any orders are obtained by constraining appropriate loadings vectors to identity vectors (vectors of ones). is the estimated array using PARAFAC model. a1-a4 (blue), b1-b4 (red) and c1-c4 (green) are loadings vectors of PARAFAC model. Unity loadings are shown by colour filled vectors and non-unity loadings by open vectors. is the estimated response at the ith instance of factor F1, jth instance of factor F2 and kth instance of factor F3 using a GEMANOVA model with a constant term (m), a main effect of F3 (c2_k), a second-order interaction between F1 and F3 (a3_ic3_k) and a third-order interaction among F1, F2 and F3 (a4_ib4_jc4_k).

Analysis of heavy metals using peptide modified sensors

Data from sensor arrays can be arranged into three-way tensors of size (sample x electrochemical signal x electrode). We investigate the application of multi-way analysis for simultaneous determination of heavy metals using peptide modified sensor arrays.

Schematic presentation of the voltamograms in (a) multi-electrode system (b) single electrode.

See our paper for more details.

Screening organometallic catalysts

The routine method of analysis for organometallic catalysts is NMR. We investigate the use of multi-way analysis applied to UV-Vis spectra as a replacement for NMR in high throughput analysis of catalysts.

Schematic presentation of a three-way PARAFAC model in which an array of X (I x J xK) is decomposed into three loadings of A (I x R), B (J x R) and C (K x R) representing relative concentration, spectral and time profiles of R components, plus a model error, E (I x J x K).

See my Ph.D. thesis for more information.