Multi-level, hybrid models and simulations, among other methods, are essential to enable predictions and hypothesis generation in systems biology research. However, the computational complexity of these models poses a bottleneck, limiting the applicability of methodologies relying on large number of simulations, such as the Optimization via Simulation (OvS) of complex biological processes. Meta-models based on approximate surrogate models simplify multi-level simulations, maintaining accuracy while reducing computational costs. Among Artificial Neural Networks (ANNs), Long Short-Term Memory (LSTM) networks are well suited to handle sequential data, which often characterizes biological simulations. This paper presents an LSTM-based surrogate modeling approach for multi-level simulations of complex biological processes. Validation relies on the simulation of Tumor Necrosis Factor (TNF) administration to a 3T3 mouse fibroblasts tumor spheroid based on PhysiBoSS 2.0, a hybrid agent-based multi-level modeling framework. Results show that the proposed LSTM meta-model is accurate and fast compared with the simulator. In fact, it infers simulated behavior with an average relative error of 7.5%. Moreover, it is at least five orders of magnitude faster. Even considering the cost of training, this approach provides a faster, more accurate, and reusable surrogate of multi-scale simulations in computationally complex tasks, such as model-based OvS of biological processes.

Fast and Accurate LSTM Meta-modeling of TNF-induced Tumor Resistance In Vitro / Abrate, Marco P; Smeriglio, Riccardo; Bardini, Roberta; Savino, Alessandro; Di Carlo, Stefano. - ELETTRONICO. - (2024), pp. 6194-6201. (Intervento presentato al convegno 2024 IEEE International Conference on Bioinformatics and Biomedicine (BIBM) tenutosi a Lisbon (PRT) nel 03-06 December 2024) [10.1109/BIBM62325.2024.10822769].

Fast and Accurate LSTM Meta-modeling of TNF-induced Tumor Resistance In Vitro

Abrate, Marco P;Smeriglio, Riccardo;Bardini, Roberta;Savino, Alessandro;Di Carlo, Stefano
2024

Abstract

Multi-level, hybrid models and simulations, among other methods, are essential to enable predictions and hypothesis generation in systems biology research. However, the computational complexity of these models poses a bottleneck, limiting the applicability of methodologies relying on large number of simulations, such as the Optimization via Simulation (OvS) of complex biological processes. Meta-models based on approximate surrogate models simplify multi-level simulations, maintaining accuracy while reducing computational costs. Among Artificial Neural Networks (ANNs), Long Short-Term Memory (LSTM) networks are well suited to handle sequential data, which often characterizes biological simulations. This paper presents an LSTM-based surrogate modeling approach for multi-level simulations of complex biological processes. Validation relies on the simulation of Tumor Necrosis Factor (TNF) administration to a 3T3 mouse fibroblasts tumor spheroid based on PhysiBoSS 2.0, a hybrid agent-based multi-level modeling framework. Results show that the proposed LSTM meta-model is accurate and fast compared with the simulator. In fact, it infers simulated behavior with an average relative error of 7.5%. Moreover, it is at least five orders of magnitude faster. Even considering the cost of training, this approach provides a faster, more accurate, and reusable surrogate of multi-scale simulations in computationally complex tasks, such as model-based OvS of biological processes.
2024
979-8-3503-8622-6
File in questo prodotto:
File Dimensione Formato  
Fast_and_Accurate_LSTM_Meta-modeling_of_TNF-induced_Tumor_Resistance_In_Vitro.pdf

accesso riservato

Tipologia: 2a Post-print versione editoriale / Version of Record
Licenza: Non Pubblico - Accesso privato/ristretto
Dimensione 1.5 MB
Formato Adobe PDF
1.5 MB Adobe PDF   Visualizza/Apri   Richiedi una copia
Paper___BIBM_2024__.pdf

accesso aperto

Tipologia: 2. Post-print / Author's Accepted Manuscript
Licenza: Pubblico - Tutti i diritti riservati
Dimensione 1.42 MB
Formato Adobe PDF
1.42 MB Adobe PDF Visualizza/Apri
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11583/2996625