Explainable artificial intelligence is increasingly used to interpret genomic, transcriptomic, and multi-omics models in drug discovery, yet the biological meaning of those explanations remains uncertain. High-dimensional dependence, pathway redundancy, incomplete prior knowledge, cohort and platform shifts, and context-specific regulation can make an explanation appear coherent while remaining unstable, model-contingent, or biologically circular. This systematic mapping review examines how explanation methods are defined, applied, and evaluated across omics-driven pharmaceutical research. A protocol-led mapping strategy was used to organize the selected peer-reviewed literature by omics layer, drug-discovery task, model architecture, explanation family, biological object, validation practice, expert involvement, and downstream decision relevance. The synthesis indicates that technical method availability is more mature than evidence of biological credibility. Feature attribution, perturbation analysis, rule extraction, attention-based interpretation, and biologically structured neural networks can reveal model-relevant patterns, but benchmark performance, pathway familiarity, and expert agreement do not alone establish mechanism, causality, or translational value. The review proposes an evidence hierarchy in which computational faithfulness, stability, context transportability, non-circular biological coherence, experimental confirmation, and decision utility are treated as distinct requirements. It also identifies major neglected areas, including cross-platform explanation replication, prospective experimental use, negative evidence, uncertainty communication, and controlled studies of expert decision impact. The central contribution is an evaluative map of where explanations are technically capable, scientifically plausible, weakly validated, or not yet ready to influence pharmaceutical decisions. The proposed standards are review-derived rather than empirically validated and require adaptation to specific biological systems, development stages, and decision consequences.