Karan Kabbur Hanumanthappa Manjunatha

Karan Kabbur Hanumanthappa Manjunatha

Auto-generated profile, based on publicly available information

Karan Kabbur Hanumanthappa Manjunatha

Scuola Superiore Meridionale

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
22 May 2023 in bioRxiv
Verified
Preprint
Open access
PDF file
KMK.K.H. Manjunatha
K.K.H. Manjunatha
MBMATTEO BRUZZONE
MATTEO BRUZZONE
GNG. Nicoletti
G. Nicoletti
SSSamir Suweis
Samir Suweis
Marco Dal MaschioMarco Dal Maschio
Marco Dal Maschio
Functional brain activity is supported by specific circuit wiring diagrams. When the synaptic annotation is available, the analysis of high-resolution connectomes allows for unraveling the circuit architectures. Despite the continuous technological and computational improvements, obtaining a whole-brain bauplan based on synaptic annotation remains a demanding effort. As an alternative, we present here an approach to extract an approximated brain connectome starting from libraries of single cell anatomical reconstructions aligned on the same anatomical reference brain. Our approach relies on the identification of neurite terminal nodes based on Strahler numbering of the cell morphology, and the adoption of a proximity range criterion, so as to infer, in the absence of synapse information, approximated brain network architectures. As an initial benchmark we used information theory metrics to confront our approach against a synaptically-annotated EM dataset of Drosophila melanogaster hemibrain. The comparison with our approach revealed a general agreement in the organization of the extracted connectivity modules measured in terms of Normalized Mutual Information, Adjusted Rand Index and Pearson correlation. Moreover, we show that the modules identified, along with their organization, can capture known circuit motives. We then applied this approach to a light microscopy dataset of the zebrafish larval brain composed of about 3000 neuronal skeletonizations. We show that the approximated connectome and the resulting modular organization is capable of capturing specific and topographically organized connection patterns as well as known functional circuit architectures. In conclusion, we present a scalable, from-circuit-to-brain range, approach to reveal approximated neuronal architectures supporting brain mechanisms, potentially suitable for hypothesis generation and for guiding the exploration of and integration with EM connectomes. Author Summary: Understanding how the brain works requires detailed maps of its neural connections, known as connectomes. While advanced techniques like electron microscopy (EM) can map these connections at the level of individual synapses, they are time consuming and resource intensive, limiting their use. As an alternative, we developed a computational method that approximates brain connectivity using existing datasets of neuron shapes (morphologies) without requiring synaptic annotations. Our approach identifies potential connections between neurons based on the proximity of their terminal branches regions where synapses are likely to form within a shared 3D reference brain. We validated our method using a high-resolution EM connectome of the fruit fly brain, demonstrating that it captures broad organizational patterns, such as clusters of densely interconnected neurons, despite moderate agreement at the synaptic level. Applying the method to a light microscopy dataset of the zebrafish larva brain (∼3,000 neurons), we successfully reconstructed large-scale networks that recapitulated known functional circuits. This approach offers a scalable way to extract brain-wide connectivity principles from existing datasets, bridging the gap between cellular anatomy and circuit function. It can guide targeted experiments and complement future EM studies, making neuroscience data more accessible for hypothesis generation.

5 June 2024
Verified
Conference Paper
Open access
JVJuan Manuel Matalobos Veiga
Juan Manuel Matalobos Veiga
RCRegino Criado
Regino Criado
MRMiguel Romance
Miguel Romance
SISergio Iglesias-Perez
Sergio Iglesias-Perez
APAlberto Partida
Alberto Partida
KMKaran Kabbur Hanumanthappa Manjunatha
Karan Kabbur Hanumanthappa Manjunatha
The Threat Intelligence discipline provides cyberdefenders with a high amount of information related to techniques, tactics and procedures involved in known cyberincidents. This information is key for the definition of a sound cybersecurity strategy, through a thorough process of risk analysis. The increasing amount of information available about known adversaries, campaigns, attacks, techniques, tactics, procedures and tools, among others, related to known cyberincidents makes it difficult to analyze and obtain useful conclusions that can be applicable to the cybersecurity strategy. Processing this information without automation leads to lack of escalability and introduction of biases and inconsistencies due to the participation of different experts with different criteria and background. The availability of a similarity measure between different cyberincidents based on their related data can be useful in this context, both increasing the efficiency identifying the cyberincidents whose analysis can provide more information, because of their dissimilitude and increasing the quality of the conclusions providing quantitative information in trends. In order to automate the processing of threat intelligence information, graphs theory can provide the capability of graphs to model and manipulate high quantities of heterogeneous data with complex relationships among them. For the definition of the similarity measure, hypergraphs can be used to model this high amount of complex and heterogeneous information along with the complex relationships between the different pieces of data, such as techniques used in attacks against banks or adversaries operating in 2023. All items related to a certain cyberincident, like adversaries, used techniques and so on, can be modelled using a hypergraphs whose nodes define the attributes that the item can or cannot have. The items related to different incidents will define different hyperedges depending on their attributes. With the model properly established, the mathematical tools available to operate with hypergraphs can be used to obtain actionable information. There are multiple similarity measures defined over hypergraphs that can be used in isolation or in combination in order to obtain similarity measures in different ways to respond to different questions like what are the techniques that have started to be used against a certain target or how similar are the attacks performed by two adversaries that can threaten a certain company.

29 January 2026 in Physical Review Research
Verified
Journal Article
Open access
JTJoakim Vianney Ngamsa Tegnitsap
Joakim Vianney Ngamsa Tegnitsap
PKPaul Didier Kamdem Kuate
Paul Didier Kamdem Kuate
KMKaran K. H. Manjunatha
Karan K. H. Manjunatha
MZManyu Zhao
Manyu Zhao
Federico D'AgataFederico D'Agata
Federico D'Agata
LMLudovico Minati
Ludovico Minati
Understanding how neurodegenerative diseases disrupt brain dynamics requires models capable of capturing the interplay between network structure and functional activity. This study presents a hardware emulation of progressive disconnection, implemented using a physical network of transistor-based chaotic electronic oscillators inspired by a biological neuronal culture. We evaluate several node disruption scenarios under different coupling strengths. These included single-node disruption and progressive damage sequences from core to periphery and periphery to core. The changes in network and node-level dynamics are assessed through the evaluation of synchronization, entropy, spectral content, and graph-theoretic properties. These analyses show that the network’s response to progressive alteration depends strongly on its topology and the coupling strength. The single-node disruption scenario revealed a high level of network resilience. Changes in average phase synchronization and topology remained modest in this scenario, consistent with functional imaging and lesion studies showing early compensatory mechanisms in healthy and mildly impaired brains. While the periphery-to-core scenario causes gradual and delayed deterioration, the core-to-periphery scenario leads to a sharp breakdown in synchronization, as well as structural and functional graph-based metrics. These profiles of degradation are comparable to those observed in neurodegenerative diseases and, in particular, Alzheimer’s disease, characterized by marked damage to central regions. Remarkably, it is found that modularity reaches a peak and then decreases just before the network collapses, a pattern that may be considered a meaningful marker of network degradation. Spectral analyses reveal a progressive shift toward low frequencies and fragmentation of the oscillatory dynamics, characterized by a decrease in high-frequency power and emergence of local frequency peaks. This study provides insights into how topology-dependent degeneration shapes brain rhythms and functional connectivity. It demonstrates that physical oscillator networks are suitable for emulating the progressive breakdown of neural circuits. The obtained insights could form the basis for hypothetical biomarkers.

1 March 2025 in Biophysics Reviews
Verified
Journal Article
Open access
EMEdoardo Milanetti
Edoardo Milanetti
KMKaran K. H. Manjunatha
Karan K. H. Manjunatha
Giancarlo RuoccoGiancarlo Ruocco
Giancarlo Ruocco
AMAmos Maritan
Amos Maritan
MFMonika Fuxreiter
Monika Fuxreiter
A wide range of higher-order structures, including dense, liquid-like assemblies, serve as key components of cellular matter. The molecular language of how protein sequences encode the formation and biophysical properties of biomolecular condensates, however, is not completely understood. Recent notion on the scale invariance of the cluster sizes below the critical concentration for phase separation suggests a universal mechanism, which can operate from oligomers to non-stoichiometric assemblies. Here, we propose a model for collective interactions in condensates, based on context-dependent variable interactions. We provide the mathematical formalism, which is capable of describing growing dynamic clusters as well as changes in their material properties. Furthermore, we discuss the consequences of the model to maximize sensitivity to the environmental signals and to increase correlation lengths.

24 June 2026 in Chaos, Solitons & Fractals
Verified
Journal Article
Closed access
KMKaran K.H. Manjunatha
Karan K.H. Manjunatha
PKPaul Didier Kamdem Kuate
Paul Didier Kamdem Kuate
JTJoakim Vianney Ngamsa Tegnitsap
Joakim Vianney Ngamsa Tegnitsap
FZFeng Zhu
Feng Zhu
YDYafang Dong
Yafang Dong
LMLudovico Minati
Ludovico Minati

24 June 2026 in Physical Review Research
Verified
Journal Article
Open access
KMKaran K. H. Manjunatha
Karan K. H. Manjunatha
JTJoakim Vianney Ngamsa Tegnitsap
Joakim Vianney Ngamsa Tegnitsap
MZManyu Zhao
Manyu Zhao
LGLucia Valentina Gambuzza
Lucia Valentina Gambuzza
YDYafang Dong
Yafang Dong
LMLudovico Minati
Ludovico Minati
This study investigates the emergence of cluster synchronization in a biologically inspired neuronal network composed of 59 coupled chaotic electronic oscillators that is modeled after a dissociated rat cortical culture. A recently introduced spectral block method is employed to predict the formation and sequence of synchronization clusters based solely on the network's structural connectivity, although the exact transition points depend on the values at which the master stability function changes sign. The analysis is conducted through numerical simulations of Rössler oscillators, which allow for a controlled numerical investigation, followed by a second part devoted to the experimental study of a circuit board composed of single-transistor chaotic oscillators. In both cases, the results confirm the expected hierarchical dynamical transition from incoherence to global synchronization via an intermediate cluster state. Furthermore, the analysis of the network of single-transistor chaotic units provides the first experimental validation of the spectral block method and demonstrates that cluster synchronization arises naturally in networks without intentionally designed symmetries. The experimental platform, which features a scalable and reconfigurable electronic network, thus bridges theoretical predictions and real-world observations, highlighting the robustness of cluster synchronization in the presence of nonidealities, noise, and parameter variations, while also emphasizing the practical considerations required when applying the method to real experiments, such as robust cluster identification and calibration of the critical points of the master stability function. These findings advance our understanding of self-organized synchronization in complex systems and have important implications for neuroscience, adaptive computing, and network engineering.