The paper introduces TANGLE, a transcriptomics-guided slide representation learning framework, which leverages gene expression profiles to guide the learning of slide embeddings from whole-slide images (WSIs). TANGLE employs a multimodal pre-training strategy that aligns modality-specific encoders using contrastive learning. The model is trained on large cohorts of publicly available (slide-expression) pairs from human and rat tissues, covering liver, breast, and lung. TANGLE demonstrates superior performance in few-shot classification, prototype-based classification, and slide retrieval tasks compared to supervised and self-supervised baselines. The study highlights the potential of transcriptomics-guided pre-training for computational pathology and provides insights into the interpretability of the learned representations.The paper introduces TANGLE, a transcriptomics-guided slide representation learning framework, which leverages gene expression profiles to guide the learning of slide embeddings from whole-slide images (WSIs). TANGLE employs a multimodal pre-training strategy that aligns modality-specific encoders using contrastive learning. The model is trained on large cohorts of publicly available (slide-expression) pairs from human and rat tissues, covering liver, breast, and lung. TANGLE demonstrates superior performance in few-shot classification, prototype-based classification, and slide retrieval tasks compared to supervised and self-supervised baselines. The study highlights the potential of transcriptomics-guided pre-training for computational pathology and provides insights into the interpretability of the learned representations.