Biotechnology equity markets exhibit extreme event-driven volatility, particularly around pivotal regulatory milestones such as U.S. Food and Drug Administration (FDA) drug approval announcements, making timing-based investment strategies both highly risky and potentially highly rewarding. This work investigates whether recurrent deep learning architectures can systematically learn latent temporal patterns in stock price behavior surrounding such regulatory events to generate profitable trading signals. A curated dataset of 177 historical FDA drug approval events was constructed, consisting of aligned pre- and post-announcement stock price time series augmented with a suite of lead technical indicators as exogenous features. A Long Short-Term Memory (LSTM) neural network was trained to perform sequence-to-signal classification for optimal buy and sell decision windows, achieving an overall predictive accuracy of approximately 87%. The results demonstrate the capacity of deep recurrent models to capture nonlinear, event-conditioned market dynamics in high-volatility biomedical equities, and suggest practical applicability for data-driven trading strategies in regulatory-sensitive financial environments.