Complex-valued representations of time-series gene expression profiles for network analysis

Background Time-series RNA sequencing provides a framework for studying dynamic gene regulation, yet conventional analyses usually represent gene expression profiles as real-valued vectors and quantify similarity using correlation or distance. Inspired by quantum information theory, we present a framework for encoding time-series gene expression profiles as complex-valued vectors comprising amplitude and phase components. Results We designed multiple encoding models that represent gene expression through the amplitude components of complex-valued vectors, encode temporal differences in the phase components, and extend the phase representation to incorporate the direction of local expression changes. Gene–gene similarity was then quantified using fidelity, which measures the overlap between two encoded vectors. Comparisons of pairwise fidelity values across the encoding models and with conventional correlation measures showed that the models produced related but distinct similarity structures. We then constructed gene–gene networks from pairwise fidelity values and detected communities of genes with similar temporal profiles. Evaluation using simulated time-series RNA-seq datasets showed moderate agreement between detected communities and the ground truth temporal patterns, with no fidelity-based model consistently outperforming weighted gene co-expression network analysis (WGCNA) or fuzzy c-means clustering across the adjusted Rand index, adjusted mutual information, pairwise F1 score, pairwise precision, and pairwise recall. In real time-series RNA-seq datasets from diverse species and biological contexts, the fidelity-based communities captured major temporal expression programs while differing among encoding m​odels in their ​community structure. Functional annotations provided exploratory biological context, and the major temporal programs broadly overlapped with those identified by WGCNA and fuzzy c-means clustering. As a supplementary proof-of-concept, we simulated SWAP-test circuits on a classical computer to estimate fidelity. Under noise-aware conditions, the simulated fidelity estimates were less accurate and computationally more costly than direct classical computation. Conclusions Representing time-series gene expression profiles as complex-valued vectors and comparing them using fidelity provides a proof-of-concept framework for temporal transcriptome analysis. The results demonstrate that encoding design changes the temporal relationships emphasized by gene–gene similarity, while not establishing uniform superiority over conventional methods.

Authors

Institutions

Publication Details

Journal
BMC Bioinformatics
Published
2026-11-18
DOI
https://doi.org/10.1186/s12859-026-06776-7
Primary Topic
Gene expression and cancer classification
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Complex-valued representations of time-series gene expression profiles for network analysis

Kentaro K. Shimizu, Jun Sese, Jianqiang Sun, Wei Cao et al.
BMC Bioinformatics
Gene expression and cancer classification
article

Complex-valued representations of time-series gene expression profiles for network analysis

Kentaro K. Shimizu, Jun Sese, Jianqiang Sun, Wei Cao, Kamachi Ikumi
article en

Abstract

Background Time-series RNA sequencing provides a framework for studying dynamic gene regulation, yet conventional analyses usually represent gene expression profiles as real-valued vectors and quantify similarity using correlation or distance. Inspired by quantum information theory, we present a framework for encoding time-series gene expression profiles as complex-valued vectors comprising amplitude and phase components. Results We designed multiple encoding models that represent gene expression through the amplitude components of complex-valued vectors, encode temporal differences in the phase components, and extend the phase representation to incorporate the direction of local expression changes. Gene–gene similarity was then quantified using fidelity, which measures the overlap between two encoded vectors. Comparisons of pairwise fidelity values across the encoding models and with conventional correlation measures showed that the models produced related but distinct similarity structures. We then constructed gene–gene networks from pairwise fidelity values and detected communities of genes with similar temporal profiles. Evaluation using simulated time-series RNA-seq datasets showed moderate agreement between detected communities and the ground truth temporal patterns, with no fidelity-based model consistently outperforming weighted gene co-expression network analysis (WGCNA) or fuzzy c-means clustering across the adjusted Rand index, adjusted mutual information, pairwise F1 score, pairwise precision, and pairwise recall. In real time-series RNA-seq datasets from diverse species and biological contexts, the fidelity-based communities captured major temporal expression programs while differing among encoding m​odels in their ​community structure. Functional annotations provided exploratory biological context, and the major temporal programs broadly overlapped with those identified by WGCNA and fuzzy c-means clustering. As a supplementary proof-of-concept, we simulated SWAP-test circuits on a classical computer to estimate fidelity. Under noise-aware conditions, the simulated fidelity estimates were less accurate and computationally more costly than direct classical computation. Conclusions Representing time-series gene expression profiles as complex-valued vectors and comparing them using fidelity provides a proof-of-concept framework for temporal transcriptome analysis. The results demonstrate that encoding design changes the temporal relationships emphasized by gene–gene similarity, while not establishing uniform superiority over conventional methods.

BMC Bioinformatics
Kihara Institute for Biological Research (JP), National Agriculture and Food Research Organization (JP), University of Zurich (CH), Research Center for Agricultural Information Technology (JP), Yokohama City University (JP)
Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung, National Agriculture and Food Research Organization, Japan Society for the Promotion of Science
Gene expression and cancer classification
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon