arXiv subjects

Manikandan Narayanan

Publications and source records attributed to Manikandan Narayanan.

4 recordsLinked to original sources

arXiv2026

A Deep Learning Earth System Model Simulation of Indian Monsoon Intraseasonal and Interannual Variability

With the data-driven artificial intelligence/machine learning (AI/ML) models having demonstrated their ability to extend the prediction horizon of large-scale weather at a fraction of computational cost of numerical weather prediction models, a pertinent question is, could these models do the same for sub-seasonal to seasonal (S2S) prediction? A key challenge in developing a S2S prediction system is the requirement for a coupled ocean-atmosphere Earth system emulator that can stably simulate the observed intraseasonal and interannual variability with fidelity. In the rapidly evolving field of AI/ML weather models, such a deep learning 3D ocean-atmosphere coupled model has become available, called SamudrACE. With our interest in developing an AI/ML S2S model for Indian monsoon, here we examine the extent to which SamudrACE faithfully simulates Indian monsoon intraseasonal and interannual variability. Compared to observation, we found biases in SamudrACE's simulation of monsoon intraseasonal and interannual variability. Our systematic documentation and analyses of these biases provide a useful benchmark for improving not only SamudrACE but also coupled emulators in general and could fast track the development of a deep learning 3D global S2S prediction system.

physics.ao-ph↗

arXiv2024

Parameterized Algorithms for the Steiner Arborescence Problem on a Hypercube

Motivated by a phylogeny reconstruction problem in evolutionary biology, we study the minimum Steiner arborescence problem on directed hypercubes (MSA-DH). Given $m$, representing the directed hypercube $\vec{Q}_m$, and a set of terminals $R$, the problem asks to find a Steiner arborescence that spans $R$ with minimum cost. As $m$ implicitly represents $\vec{Q}_m$ comprising $2^{m}$ vertices, the running time analyses of traditional Steiner tree algorithms on general graphs does not give a clear understanding of the actual complexity of this problem. We present algorithms that exploit the structure of the hypercube and run in time polynomial in $|R|$ and $m$. We explore the MSA-DH problem on three natural parameters - $R$, and two above-guarantee parameters, number of Steiner nodes $p$ and penalty $q$. For above-guarantee parameters, the parameterized MSA-DH problem takes $p \geq 0$ or $q\geq 0$ as input, and outputs a Steiner arborescence with at most $|R| + p - 1$ or $m + q$ edges respectively. We present the following results ($\tilde{\mathcal{O}}$ hides the polynomial factors): 1. An exact algorithm that runs in $\tilde{\mathcal{O}}(3^{|R|})$ time. 2. A randomized algorithm that runs in $\tilde{\mathcal{O}}(9^q)$ time with success probability $\geq 4^{-q}$. 3. An exact algorithm that runs in $\tilde{\mathcal{O}}(36^q)$ time. 4. A $(1+q)$-approximation algorithm that runs in $\tilde{\mathcal{O}}(1.25284^q)$ time. 5. An $\mathcal{O}\left(p\ell_{\mathrm{max}} \right)$-additive approximation algorithm that runs in $\tilde{\mathcal{O}}(\ell_{\mathrm{max}}^{p+2})$ time, where $\ell_{\mathrm{max}}$ is the maximum distance of any terminal from the root.

cs.DS↗

arXiv2022

Using Signal Processing in Tandem With Adapted Mixture Models for Classifying Genomic Signals

Genomic signal processing has been used successfully in bioinformatics to analyze biomolecular sequences and gain varied insights into DNA structure, gene organization, protein binding, sequence evolution, etc. But challenges remain in finding the appropriate spectral representation of a biomolecular sequence, especially when multiple variable-length sequences need to be handled consistently. In this study, we address this challenge in the context of the well-studied problem of classifying genomic sequences into different taxonomic units (strain, phyla, order, etc.). We propose a novel technique that employs signal processing in tandem with Gaussian mixture models to improve the spectral representation of a sequence and subsequently the taxonomic classification accuracies. The sequences are first transformed into spectra, and projected to a subspace, where sequences belonging to different taxons are better distinguishable. Our method outperforms a similar state-of-the-art method on established benchmark datasets by an absolute margin of 6.06% accuracy.

q-bio.GN↗

arXiv2007

Comparing Protein Interaction Networks via a Graph Match-and-Split Algorithm

We present a method that compares the protein interaction networks of two species to detect functionally similar (conserved) protein modules between them. The method is based on an algorithm we developed to identify matching subgraphs between two graphs. Unlike previous network comparison methods, our algorithm has provable guarantees on correctness and efficiency. Our algorithm framework also admits quite general connectivity and local matching criteria that define when two subgraphs match and constitute a conserved module. We apply our method to pairwise comparisons of the yeast protein network with the human, fruit fly and nematode worm protein networks, using a lenient criterion based on connectedness and matching edges, coupled with a betweenness clustering heuristic. We evaluate the detected conserved modules against reference yeast protein complexes using sensitivity and specificity measures. In these evaluations, our method performs competitively with and sometimes better than two previous network comparison methods. Further under some conditions (proper homolog and species selection), our method performs better than a popular single-species clustering method. Beyond these evaluations, we discuss the biology of a couple of conserved modules detected by our method. We demonstrate the utility of network comparison for transferring annotations from yeast proteins to human ones, and validate the predicted annotations.

q-bio.MN↗