Electroencephalography (EEG) is widely used in the detection of neurological disorders due to its high temporal resolution, yet the development of clinically meaningful machine learning (ML) models remains limited by the scarcity of disease-specific EEG datasets, particularly for Parkinson’s disease (PD), Alzheimer’s disease (AD), schizophrenia (SCZ), epilepsy (EPI), and frontotemporal dementia (FTD). This study proposes a biologically informed simulation framework to generate synthetic EEG data by injecting physiologically realistic disease patterns into healthy recordings. Guided by clinical literature, the framework applies frequency-specific modulations and region-targeted alterations to emulate disorder-specific neural signatures, including theta and delta amplification in PD and SCZ, alpha suppression in AD and FTD, and gamma bursts with transient spikes in EPI, using explicit mathematical formulations. Unlike generative adversarial network-based models that rely solely on data-driven synthesis, this method embeds known spectral and spatial biomarkers directly into the signals. The generated EEG data were validated through time frequency analysis, topographic mapping, and statistical testing, revealing strong spectral separability and realistic spatial and temporal features across conditions. These findings confirm the biological plausibility and discriminative strength of the proposed method, providing a scalable and clinically relevant approach for EEG dataset generation and the development of diagnostic ML models.