The aim of this study was to assess the potential health risks of radioactive elements and heavy metals ingested through the consumption of various types of tea imported to the Polish market (black, green, red, oolong and white). The concentrations [Bq/kg] of radionuclides (⁴⁰K, ¹³⁷Cs, ²²⁶Ra, ²¹⁰Pb and ²²⁸Th) in tea leaves before and after brewing were measured using γ-ray spectrometry with high-purity germanium (HPGe). The concentrations [mg/kg] of the studied elements (Fe, Cr, Cu, Mo, Al, Mn, Ni, P, V, Cd and Pb) were determined using a microwave-induced plasma optical emission spectrometer (MIP-OES). The results presented here will help to expand the database of heavy metals and radioactivity in teas. With regard to the potential health risk, the percentage of leaching of individual elements in different types of tea infusions was determined, and the assessment of the consumption risk was estimated. Since the calculated exposure factors, namely the HQ (Hazard Quotient) and THQ (Target Hazard Quotient), do not exceed critical levels, teas can still be considered health-beneficial products (most of the radionuclides as well as elements remain in the leaves (65–80%) after brewing).

The aim of the following research is to assess the applicability of calculated quantum properties of molecular fragments as molecular descriptors in machine learning classification task. The research is based on bio-concentration and QM9-extended databases. A number of compounds with results from quantum-chemical calculations conducted with Psi4 quantum chemistry package was also added to the quantum properties database. Classification results are compared with a baseline of random guesses and predictions obtained with the traditional RDKit generated molecular descriptors. Chosen classification metrics show that results obtained with fragments quantum descriptors fall between results from baseline and those provided by molecular descriptors widely applied in cheminformatics. According to the results, the implementation of principal component analysis, causes a drop in categorization metrics.

In the following research, a new modification of traditional atom pairs is studied. The atom pairs are enriched with values originating from quantum chemistry calculations. A random forest machine learning algorithm is applied to model 10 different properties and biological activities based on different molecular representations, and it is evaluated via repeated cross-validation. The predictive power of modified atom pairs, quantum atom pairs, are compared to the predictive powers of traditional molecular representations known and widely applied in cheminformatics. The root mean squared error (RMSE), R (2), area under the receiver operating characteristic curve (AUC) and balanced accuracy were used to evaluate the predictive power of the applied molecular representations. Research has shown that while performing regression tasks, quantum atom pairs provide better fits to the data than do their precursors.

The aim of the present research was to assess the radiological importance of the fly ashes generated from the incineration of sludge from the “Czajka” sewage treatment plant located in Warsaw, Poland. The γ-spectrometry technique was employed for their radiometry. The high purity germanium (HPGe) detector was mathematically energy-efficiency calibrated. The ash samples were measured two times, i.e., after 24 days and after 3 months of storage. Storage was supposed to provide a radioactive equilibrium. Radionuclides from three natural decay series, i.e., uranium–radium, uranium–actinium, and thorium, were detected in the ash samples. Apart from these, ⁴⁰ K, ¹³⁷ Cs, and ⁷ Be were also detected. The indicated levels of radionuclides do not produce radioactive hazards, regardless of the method used for further management.

The following research assesses the capability of machine learning in predicting maximum emission wavelengths of organic compounds. The predictions are based on molecular descriptors and fingerprints widely applied in cheminformatics. In an effort to further improve accuracy, developed machine learning models were enriched with quantum mechanics derived features. Multi linear, gradient boosting and random forest regressions were applied. Computers were trained and tested with database of experimental data of optical properties.