日時：	2025年7月1日（火）13：30～15：00／13:30 - 15:00, July 1 (Tue), 2025
場所：	医生物学研究所　医生研３号館５階501室 Room 501, 5th floor, Bldg. #3 Institute for Life and Medical Sciences, Kyoto University
演者：	Ian Overton Patrick G Johnston Centre for Cancer Research, Queen’s University Belfast, United Kingdom
演題：	Integrative Analytics Connecting Genotype and Phenotype for Precision Oncology

講演要旨

Understanding the molecular mechanisms that control the biology of health and disease requires
multiscale models that map relationships between genes and phenotypes. Measuring,
parameterising and simulating molecular systems in exhaustive detail is typically impossible due to
biological complexity, our limited knowledge and limited available data. Therefore, simplifying
abstractions in concert with empirical analysis of matched genome-scale and descriptive data are
valuable strategies. We developed the Gabi algorithm to connect clinical variables with molecular
measurements; including a novel relevance thresholding procedure and information-theoretic
directionality inference. Gabi outperformed existing state-of-the-art approaches on blind test data.
We applied Gabi to derive a causal information-flow network for invasive hormone-driven breast
tumours. Findings include a switch involving the estrogen receptor (ER𝛂) and its phosphorylated
form, which had opposing regulatory effects on many common targets. Gabi predicted proteins that
influence important clinical parameters (e.g. tumour stage) and Feedback Vertex Set (FVS) analysis
revealed key network control nodes. Analysis of our causal network identified patient risk groups that
have prognostic value in multivariate modeling controlling for clinical variables. I will also discuss the
NetNC algorithm [Cancers 2020;12:2823] and SynLeGG resource [Nucleic Acids Research
2021;49:W613-8]. NetNC recovers the network-defined signal in noisy data, for example defining
biologically coherent modules in matched drug-sensitive vs drug-resistant n=1 patient samples.
SynLeGG and the MultiSEp algorithm interrogate mutually exclusive loss signatures in multiomics
data, towards context-specific synthetic lethal drug targets and companion diagnostic biomarkers.
Application of MultiSEp to Multiple Myeloma (MM) patients and cytogenetic subtypes revealed
context-specific synthetic lethal networks that inform fundamental biology, including for poorly
characterised and noncoding genes. Almost all MM patients relapse and succumb to therapyresistant
disease; accordingly, more effective treatments are urgently needed. Analysis of our
predicted MM patient synthetic lethal networks reveals individual ‘nexus’ genes where the network
neighbourhood genes are collectively mutated in a relatively high proportion of MM cohorts,
representing attractive drug targets. Laboratory follow-up further validates our computational
approaches.（Language: English）

Seminar_20250701

＜本セミナーは事前申し込みが必要です。＞
参加申込 連絡先（Registration）：
Ayako Shiode
Laboratory of Mathematical Biology,
Institute for Life and Medical Sciences, Kyoto University
Email: shiode.ayako.2k[@]kyoto-u.ac.jp

contact:
Atsushi Mochizuki
Laboratory of Mathematical Biology,
Institute for Life and Medical Sciences, Kyoto University
Email: mochi[@]infront.kyoto-u.ac.jp