Despite the small percentage of protein-coding genes in the human genome, most genetic variants associated with disease are located in non-protein coding regions, hinting at the regulatory roles of long non-coding RNAs (lncRNAs). Our research leverages real-time nanopore sequencing and bioinformatics to systematically characterize lncRNA functions, focusing on leukemia. We have discovered that real-time data processing with machine learning classifiers trained on gene-level short read data is sufficient to reduce the time and cost of cancer diagnosis by two orders of magnitude. While evaluating this approach on ~20 patients, we leveraged the resulting nanopore sequencing data to assemble a 'meta-transcriptome', which improves disease subtype clustering while revealing hundreds of new (often repeat harbouring) lncRNAs, many of which are among the most differentially expressed between disease subtypes. We performed a CRISPR interference screen on 760 lncRNAs followed by smFISH assays to identify critical lncRNAs affecting cellular proliferation, mapped to specific molecular pathways. We also evaluated the latest nanopore sequencing for detecting m6A modifications in native RNA sequences, revealing differential methylation patterns that suggest novel regulatory dimensions. This work not only underscores the complexity of lncRNAs but also sets a foundation for their integration into precision medicine.