Сравнение компьютерных предикторов биологической активности органических соединений (аналитический обзор)
- № 2(8) 2017
Страницы:
76
–
80
Язык: русский
Аннотация
Как правило, в виртуальном скрининге для построения компьютерных предикторов большинство
пользователей используют QSAR-моделирование или подход на основе химического сходства в зависимости от
их опыта и/или доступности инструмента. Целью работы является сравнение этих двух основных подходов на
одном эталонном наборе данных, где прогностическая эффективность сопоставлялась бы с учетом точности
прогнозирования как на обучающих, так и на внешних наборах данных. В обзоре представлены два метода
прогноза биологической активности, реализованные в виде онлайн-пакетов, — SEA, PASS и схема KNN QSAR.
Результаты вычислительных экспериментов по этим трем подходам показали преимущество схемы KNN
QSAR. Методы, рассмотренные в работе, представляют интерес для химиков и экспериментальных биологов,
работающих в области биологического скрининга химических библиотек.
As a rule, in the virtual screening to build predictors most users use QSAR-modeling approach or approach based on chemical similarity; it’s depend on their experience and / or tools available. The purpose was to review the investigations where compared to a reference dataset of these two basic approaches, predictive effectiveness of which was compared to that based on the prediction of both the training, and external data sets. We present two methods of biological activity prediction realized in the form of on-line packages, — SEA, PASS and KNN QSAR scheme. The results of computational experiments have shown the advantage of the latter scheme. The methods discussed in the work should be useful for chemists and experimental biologists working in the field of biological screening of chemical libraries.
Odatda, Virtual skriningda ko‘pchilik foydalanuvchilar kompyuterli prediktor qurish uchun QSAR modellashtirishdan foydaniladi yoki ularning tajriba va/yoki vositalari mavjudligiga qarab kimyoviy o‘xshashlik asosida yondashiladi. Ishning maqsadi bu ikkita asosiy yondashuv bilan bitta etalondagi ma’lumotlarni solishtirish bo‘yicha ishlarning tavsiflanishi, bunda o‘rganilayotgan va tashqi ma’lumotlar to‘plamlaridagi bashoratlash aniqligini hisobga olish bilan bashoratlash samaradorligi taqqoslandi. SEA, PASS va KNN QSAR- on-layn paketlar ko‘rinishida amalga oshirilib, biologik aktivlikni prognoz qiladigan ikkita metod taqdim etilgan. Hisoblash natijalari ikkinchi sxema afzalligini ko‘rsatdi. Mazkur ishda qaralgan metodlar kimyoviy biblotekalarni biologik skriningi sohasida ishlaydigan kimyogarlar va eksprimental biologlarga foydali bo‘lishi kerak.
As a rule, in the virtual screening to build predictors most users use QSAR-modeling approach or approach based on chemical similarity; it’s depend on their experience and / or tools available. The purpose was to review the investigations where compared to a reference dataset of these two basic approaches, predictive effectiveness of which was compared to that based on the prediction of both the training, and external data sets. We present two methods of biological activity prediction realized in the form of on-line packages, — SEA, PASS and KNN QSAR scheme. The results of computational experiments have shown the advantage of the latter scheme. The methods discussed in the work should be useful for chemists and experimental biologists working in the field of biological screening of chemical libraries.
Odatda, Virtual skriningda ko‘pchilik foydalanuvchilar kompyuterli prediktor qurish uchun QSAR modellashtirishdan foydaniladi yoki ularning tajriba va/yoki vositalari mavjudligiga qarab kimyoviy o‘xshashlik asosida yondashiladi. Ishning maqsadi bu ikkita asosiy yondashuv bilan bitta etalondagi ma’lumotlarni solishtirish bo‘yicha ishlarning tavsiflanishi, bunda o‘rganilayotgan va tashqi ma’lumotlar to‘plamlaridagi bashoratlash aniqligini hisobga olish bilan bashoratlash samaradorligi taqqoslandi. SEA, PASS va KNN QSAR- on-layn paketlar ko‘rinishida amalga oshirilib, biologik aktivlikni prognoz qiladigan ikkita metod taqdim etilgan. Hisoblash natijalari ikkinchi sxema afzalligini ko‘rsatdi. Mazkur ishda qaralgan metodlar kimyoviy biblotekalarni biologik skriningi sohasida ishlaydigan kimyogarlar va eksprimental biologlarga foydali bo‘lishi kerak.