Biology subjects

Ku, M.

Publications and source records attributed to Ku, M..

3 recordsLinked to original sources

bioRxiv2026

MHC Attention: Identifying HLA-E presented cancer antigens through deep learning and high-throughput screening

HLA-E presented cancer peptides can be promising cancer therapy targets, as HLA-E is minimally polymorphic and widely expressed across human populations and cancer types. However, systematic discovery of cancer associated HLA-E peptides has been constrained by sparse training data and the technical difficulty of HLA-E immunopeptidomics. Here we develop an integrated HLA-E antigen discovery platform combining a deep learning prediction model, pooled mammalian cell screening, peptide-HLA-E stability validation, and mass spectrometry. We introduce MHC Attention, a neural network that learns allele-level attention over candidate MHC alleles in multi-allele immunopeptidomics datasets, enabling direct training on patient-derived MHC peptide data. Screening an approximately 6,000-peptide HLA-E library identified stable HLA-E-presented peptides and generated HLA-E-specific training data that improved prediction performance of MHC Attention. Combining our screening assays and improved prediction algorithm, we discovered novel HLA-E-presented cancer peptides, including candidates derived from ETV4, WT1, RNF43 and BMP8A, with orthogonal support from stability assays or immunopeptidomics. These results establish a scalable framework for HLA-E peptide target discovery and provide candidate targets for broadly applicable peptide-HLA-directed cancer immunotherapies. MHC Attention 2.0 can be accessed online via https://vcreate.io/mhcattention.

immunology↗

bioRxiv2026

Existence of Causation without Correlation in Transcriptional Networks

It is commonly assumed that lack of correlation is evidence for lack of causal relationship. Here, however we show that in transcriptional networks, causal linkages can exist in the absence of correlation. We find that a substantial proportion of transcribed genes in yeast and in mouse, show evidence of state-dependent (nonlinear) and temporally coordinated dynamics in their expression patterns (65-77%). Using a test that accommodates this fact, we uncover strong causal relationships that are invisible to correlation-based analyses for both yeast and mouse models. Specifically, for yeast we detect uncorrelated causal relationships for the transcriptional regulators WHI5 and YHP1, and can verify these relationships experimentally. These genes reside at important checkpoints in the cell cycle where multiple signals are integrated at single nodes, giving rise to causal relationships, that despite being uncorrelated, can be accurately detected (71-78%) using a nonlinear causality test.

systems biology↗

bioRxiv2026

Kv11.1 (hERG) Protein Interaction Networks Connect Endocytic Trafficking to Polygenic Influences on Cardiac Repolarization

Polygenic scores (PGS) summarize the combined effects of common single-nucleotide polymorphisms and contribute to predictions of disease severity, but biological consequences linked to these common variants remain poorly defined. Here, we focused on polygenic liability for a measurable electrophysiologic trait (the QT interval). Prolonged QT interval, measured on patient electrocardiograms, is associated with an increased risk for cardiac arrhythmia. We investigated human induced pluripotent stem cell cardiomyocytes (hiPSC-CMs) from donors with extreme PGS (i.e., high and low) related to QT interval duration. We paired global proteomics with multiplexed affinity purification mass spectrometry (AP-MS) centered on Kv11.1 (hERG), a major determinant of QT-interval repolarization. Global proteomics indicated increased mitochondrial protein abundance in high-PGS cardiomyocytes, but this did not explain the Kv11.1 interactome. In high-PGS cells, Kv11.1 showed increased associations with myosin motor proteins and endosomal recycling machinery, consistent with altered (and potentially increased) recycling/trafficking dynamics rather than trafficking deficiency observed with most pathogenic Kv11.1 variants. This proof-of-concept study underscores a framework for linking polygenic factors to tractable biological consequences by combining patient-specific hiPSCs, proteomics and affinity-purification. Linking polygenic scores to changes in protein networks provides testable mechanisms that can be applied across many diseases. Significance StatementPolygenic scores (PGS) predict disease risk, but how biological pathways are influenced by these common variants remains difficult to define. We generated human induced pluripotent stem cells from individuals with extreme high- and low-PGS for QT interval, a key electrocardiographic measure linked to arrhythmia risk. By combining global proteomics and interactomics for a common ion channel involved in regulating the QT interval (Kv11.1) we found mechanisms that are influenced by common genetic traits in patients. Our work connects polygenic scores to pathway-level molecular mechanisms in human cells and provides a general framework for uncovering how complex genetic architecture drives disease-relevant biology. ClassificationBiological Sciences; Medical Sciences

genomics↗