|
|
| Type |
Poster Presentation |
| Area |
Analytical Chemistry |
| Room No. |
Event Hall |
| Time |
4월 19일 (목요일) 11:00~12:30 |
| Code |
ANAL.P-234 |
| Subject |
Development of Screening Software for Illicit Drugs and Analogues and Classification of Erectile Dysfunction Drugs |
| Authors |
Inae Jang, Han Bin Oh*
Department of Chemistry, Sogang University, Korea
|
| Abstract |
Illicit drugs and analogues, for example, erectile dysfunction (ED) drugs, analgesics, diuretics, weight loss compounds, and psychotropic drugs, are widely spread in the online markets, particularly advertised as health supplements. To monitor these illegal drugs and analogues from the illegal market places, a variety of analytical screening tools are now being used, and LC-MS/MS-based method has been recognized as one of the most powerful screening tool. For screening these illegal drugs, LC-MS/MS screening software, coded with MATLAB, was developed, wherein in-silico expanded database was used. A screening software, named as ‘Spectra Match’, consists of two layers. The first layer is the viewers for chromatogram and mass spectra. This software layer was designed to upload mzxml input data files and a variety of chromatogram formats can be extracted from raw data. The first layer is coupled with the second layer in which the identification of compounds under LC-MS/MS investigation is a main function. The database has about 50,000 entries, including in-silico MS/MS spectra for a large number of ED analogues. Secondly, as an additional option, a machine learning-based classification model is included to classify unknown ED drug analogues in the software. Capturing the features for the LC-MS/MS spectra of ED drug analogues was made by machine learning methods. For this the LC-MS/MS spectra for ED drug analogues were converted to binary bar-code spectra, and these bar-code spectra were trained with various machine learning techniques such as CART, random forest, KNN, SVM and ANN. Bar-code tandem mass spectra of the ED samples categorized into four groups i.e., tadalafil, sildenafil, vardenafil group, and the others, based on their structural similarities were machine-learned and a multiclass classification statistical model was constructed. It was found that the ANN model could classify the tandem mass spectra of unknown samples (20% external test set) with 100 % accuracy. This machine-learned ANN model was included in the software as a screening tool for ED drug analogues that may be illicitly added into health supplements.
This research was supported by a grant (18182MFDS425) from Ministry of Food and Drugs Safety of Korea
|
| E-mail |
musiccat10@naver.com |
|
121st General Meeting of the KCS