arXiv subjects

Alessio Fallani

Publications and source records attributed to Alessio Fallani.

5 recordsLinked to original sources

arXiv2026

Data-Efficient Active Learning Discovery of Transition Metal Photosensitizers for Type I Photodynamic Therapy

Transition-metal complexes (TMCs) are promising photosensitizers for Type~I photodynamic therapy (PDT), where electron-transfer processes can generate reactive oxygen species under hypoxic conditions. Yet identifying candidates with the required ground- and excited-state redox energetics remains challenging across the vast chemical space of TMCs. Here, we develop a data-efficient active learning (AL) framework for the discovery of Type~I active TMC photosensitizers by combining a chemically structured design space of over 2.1 million Ru(II), Os(II), and Ir(III) complexes with targeted DFT calculations and pretrained atomistic representations. With only 300 quantum-chemical evaluations, the approach efficiently enriches candidates within a mechanistically defined optimal redox region. Analysis of the viable complexes reveals chemical design principles linking metal identity, ligand framework, substituent pattern, and physicochemical properties to Type~I photoreactivity, including a pronounced preference for Os(II)-based complexes and electronically asymmetric ligand environments along with combination of electronic donating and accepting substituents. More broadly, the strategy presented herein provides a scalable, mechanism-guided route for the rational design of transition-metal photocatalysts for applications spanning biomedicine, solar energy conversion, and photoredox chemistry.

physics.chem-ph↗

arXiv2024

Pretraining Graph Transformers with Atom-in-a-Molecule Quantum Properties for Improved ADMET Modeling

We evaluate the impact of pretraining Graph Transformer architectures on atom-level quantum-mechanical features for the modeling of absorption, distribution, metabolism, excretion, and toxicity (ADMET) properties of drug-like compounds. We compare this pretraining strategy with two others: one based on molecular quantum properties (specifically the HOMO-LUMO gap) and one using a self-supervised atom masking technique. After fine-tuning on Therapeutic Data Commons ADMET datasets, we evaluate the performance improvement in the different models observing that models pretrained with atomic quantum mechanical properties produce in general better results. We then analyse the latent representations and observe that the supervised strategies preserve the pretraining information after finetuning and that different pretrainings produce different trends in latent expressivity across layers. Furthermore, we find that models pretrained on atomic quantum mechanical properties capture more low-frequency laplacian eigenmodes of the input graph via the attention weights and produce better representations of atomic environments within the molecule. Application of the analysis to a much larger non-public dataset for microsomal clearance illustrates generalizability of the studied indicators. In this case the performances of the models are in accordance with the representation analysis and highlight, especially for the case of masking pretraining and atom-level quantum property pretraining, how model types with similar performance on public benchmarks can have different performances on large scale pharmaceutical data.

cs.LG↗

arXiv2023

Enabling Inverse Design in Chemical Compound Space: Mapping Quantum Properties to Structures for Small Organic Molecules

Computer-driven molecular design combines the principles of chemistry, physics, and artificial intelligence to identify novel chemical compounds and materials with desired properties for a specific application. In particular, quantum-mechanical (QM) methods combined with machine learning (ML) techniques have accelerated the estimation of accurate molecular properties, providing a direct mapping from 3D molecular structures to their properties. However, the development of reliable and efficient methodologies to enable \emph{inverse mapping} in chemical space is a long-standing challenge that has not been accomplished yet. Here, we address this challenge by demonstrating the possibility of parametrizing a given chemical space with a finite set of extensive and intensive QM properties. In doing so, we develop a proof-of-concept implementation that combines a Variational Auto-Encoder (VAE) trained on molecular structures with a property encoder designed to learn the latent representation from a set of QM properties. The result of this joint architecture is a common latent space representation for both structures and properties, which enables property-to-structure mapping for small drug-like molecules contained in the QM7-X dataset. We illustrate the capabilities of our approach by conditional generation of \emph{de novo} molecular structures with targeted properties, transition path interpolation for chemical reactions as well as insights into property-structure relationships. Our findings thus provide a proof-of-principle demonstration aiming to enable the inverse property-to-structure design in diverse chemical spaces.

physics.chem-ph↗

arXiv2023

Modeling Non-Covalent Interatomic Interactions on a Photonic Quantum Computer

Non-covalent interactions are a key ingredient to determine the structure, stability, and dynamics of materials, molecules, and biological complexes. However, accurately capturing these interactions is a complex quantum many-body problem, with no efficient solution available on classical computers. A widely used model to accurately and efficiently model non-covalent interactions is the Coulomb-coupled quantum Drude oscillator (cQDO) many-body Hamiltonian, for which no exact solution is known. We show that the cQDO model lends itself naturally to simulation on a photonic quantum computer, and we calculate the binding energy curve of diatomic systems by leveraging Xanadu's Strawberry Fields photonics library. Our study substantially extends the applicability of quantum computing to atomistic modeling, by showing a proof-of-concept application to non-covalent interactions, beyond the standard electronic-structure problem of small molecules. Remarkably, we find that two coupled bosonic QDOs exhibit a stable bond. In addition, our study suggests efficient functional forms for cQDO wavefunctions that can be optimized on classical computers, and capture the bonded-to-noncovalent transition for increasing interatomic distances. Remarkably, we find that two coupled bosonic QDOs exhibit a stable bond. In addition, our study suggests efficient functional forms for cQDO wavefunctions that can be optimized on classical computers, and capture the bonded-to-noncovalent transition for increasing interatomic distances.

quant-ph↗

arXiv2021

Learning feedback control strategies for quantum metrology

We consider the problem of frequency estimation for a single bosonic field evolving under a squeezing Hamiltonian and continuously monitored via homodyne detection. In particular, we exploit reinforcement learning techniques to devise feedback control strategies achieving increased estimation precision. We show that the feedback control determined by the neural network greatly surpasses in the long-time limit the performances of both the "no-control" strategy and the standard "open-loop control" strategy, which we considered as benchmarks. We indeed observe how the devised strategy is able to optimize the nontrivial estimation problem by preparing a large fraction of trajectories corresponding to more sensitive quantum conditional states.

quant-ph↗