This data set, referenced as "toy data set" in the article "A deep-learning based analysis framework for ultra-high throughput screening time-series data" (https://doi.org/10.1101/2024.08.22.609110), mimics a high-throughput screening data set. To demonstrate the application of our analysis framework described in the main article this toy data set was generated. It contains in total 1536000 individual transient signals, splitted in 5 batches of each 200 plates in 1536-well plate format. Five distinct signal classes were used to resemble typical shapes encountered in biological experiments. Fequency of occurrences for each class is reported in the main article.