arXiv subjects

Yahya Forghani

Publications and source records attributed to Yahya Forghani.

4 recordsLinked to original sources

arXiv2026

Beyond Network Topology: Biological Evidence Integration and Reproducible Benchmarking for Protein Complex Detection

Protein complexes are molecular assemblies that coordinate cellular regulation, signaling, metabolism, and disease-relevant protein function. Detecting such assemblies from protein-protein interaction (PPI) networks remains challenging because network topology is an incomplete abstraction: an edge may represent direct binding, functional association, co-complex evidence, co-expression, co-localization, or a computationally predicted interaction. This focused critical methodological review examines how biological evidence can improve protein-complex detection beyond dense-subgraph discovery. We consider Gene Ontology, expression, localization, domains and motifs, sequence and structure, interface evidence, temporal context, RNA or regulatory evidence, and representation learning, while retaining classical graph-clustering methods as historical baselines. Interpretable evidence-aware graph methods currently provide a strong balance between biological plausibility and reproducibility, whereas structure-aware, temporal, heterogeneous, and hypergraph models offer greater biological realism but require stronger independent benchmarking. Reported F-measures cannot be directly compared across incompatible PPI releases, reference sets, matching thresholds, preprocessing pipelines, and metric implementations. Progress therefore requires fixed dataset versions, explicit controls for Gene Ontology circularity, overlap-aware metrics, uncertainty estimates, and executable software packages. Reliable protein-complex detection ultimately depends on connecting graph-based predictions to molecular structure, interaction mechanisms, cellular context, and functional assembly.

cs.SI↗

arXiv2026

ECHO-PPI: Evidence-Bundled Overlapping Protein Module Detection with Hierarchical Assignment Confidence for Network Biology

Identifying protein modules in protein-protein interaction (PPI) networks is central to understanding cellular organisation, yet many community-detection methods treat module membership as a binary output with limited assignment-level justification. Proteins that participate in multiple complexes as shared subunits, peripheral interactors, or context-dependent bridges can be overlooked when networks are forced into hard partitions, and even overlapping methods rarely provide traceable evidence for individual protein-module assignments. We present ECHO-PPI, a framework for overlapping protein-module detection that combines competitive module discovery with structured assignment-level interpretation. For every protein-module assignment, ECHO-PPI exports an evidence bundle combining weighted topology, semantic functional similarity, Gene Ontology support, provenance fields, and hierarchical confidence labels: Core, Inner, Outer, and Uncertain. This makes each assignment inspectable and reproducible rather than an opaque membership claim. We benchmark ECHO-PPI on two yeast PPI resources, the Gavin socioaffinity network and the Krogan 2006 dataset, against MCL, MCL+overlap, ClusterONE, and SLPA. ECHO-PPI achieves predictive parity with overlap-aware baselines while being the only evaluated method to provide complete required-field evidence bundles. Core assignments show the strongest gold-standard support and consistent multi-channel evidence across both datasets. By separating predictive clustering from evidence-bundled interpretation, ECHO-PPI provides computational biologists with a path from cluster lists to defensible, reproducible protein-module hypotheses suitable for curator-facing network-biology workflows.

cs.SI↗

arXiv2020

Distance metric learning based on structural neighborhoods for dimensionality reduction and classification performance improvement

Distance metric learning can be viewed as one of the fundamental interests in pattern recognition and machine learning, which plays a pivotal role in the performance of many learning methods. One of the effective methods in learning such a metric is to learn it from a set of labeled training samples. The issue of data imbalance is the most important challenge of recent methods. This research tries not only to preserve the local structures but also covers the issue of imbalanced datasets. To do this, the proposed method first tries to extract a low dimensional manifold from the input data. Then, it learns the local neighborhood structures and the relationship of the data points in the ambient space based on the adjacencies of the same data points on the embedded low dimensional manifold. Using the local neighborhood relationships extracted from the manifold space, the proposed method learns the distance metric in a way which minimizes the distance between similar data and maximizes their distance from the dissimilar data points. The evaluations of the proposed method on numerous datasets from the UCI repository of machine learning, and also the KDDCup98 dataset as the most imbalance dataset, justify the supremacy of the proposed approach in comparison with other approaches especially when the imbalance factor is high.

cs.LG↗

arXiv2013

Comment on "robustness and regularization of support vector machines" by H. Xu, et al., (Journal of Machine Learning Research, vol. 10, pp. 1485-1510, 2009, arXiv:0803.3490)

This paper comments on the published work dealing with robustness and regularization of support vector machines (Journal of Machine Learning Research, vol. 10, pp. 1485-1510, 2009) [arXiv:0803.3490] by H. Xu, etc. They proposed a theorem to show that it is possible to relate robustness in the feature space and robustness in the sample space directly. In this paper, we propose a counter example that rejects their theorem.

cs.LG↗