arXiv subjects

Diogo Soares

Publications and source records attributed to Diogo Soares.

4 recordsLinked to original sources

arXiv2026

OmegAMP: Targeted AMP Discovery via Biologically Informed Generation

Deep learning-based antimicrobial peptide (AMP) discovery faces critical challenges such as limited controllability, lack of representations that efficiently model antimicrobial properties, and low experimental hit rates. To address these challenges, we introduce OmegAMP, a framework designed for reliable AMP generation with increased controllability. Its diffusion-based generative model leverages a novel conditioning mechanism to achieve fine-grained control over desired physicochemical properties and to direct generation towards specific activity profiles, including species-specific effectiveness. This is further enhanced by a biologically informed encoding space that significantly improves overall generative performance. Complementing these generative capabilities, OmegAMP leverages a novel synthetic data augmentation strategy to train classifiers for AMP filtering, drastically reducing false positive rates and thereby increasing the likelihood of experimental success. Our in silico experiments demonstrate that OmegAMP delivers state-of-the-art performance across key stages of the AMP discovery pipeline, enabling us to achieve an unprecedented success rate in wet lab experiments. We tested 25 candidate peptides, 24 of them (96%) demonstrated antimicrobial activity, proving effective even against multi-drug resistant strains. Our findings underscore OmegAMP's potential to significantly advance computational frameworks in the fight against antimicrobial resistance.

cs.LG↗

arXiv2026

Scalable and Interpretable Representation Alignment with Ordinal Similarity

Evaluating representation similarity is fundamental to representation learning. However, existing metrics suffer from significant limitations: they lack interpretability due to shifting baselines, lack robustness to outliers, and are computationally intractable for large datasets, forcing reliance on heuristic approximations. To address this, we develop an ordinal-similarity framework, instantiated by the Triplet (TSI) and Quadruplet (QSI) Similarity Indices, which measure alignment by quantifying the consistency of ordinal relationships. We theoretically demonstrate this formulation is inherently interpretable, robust to outliers, and computationally efficient. Finally, we establish a formal equivalence between TSI and local neighborhood alignment, measured by Mutual Nearest Neighbors. Empirically, we validate these properties and show that ordinal similarity offers a scalable approach to measuring alignment, enabling practitioners to better understand and design representations.

cs.LG↗

arXiv2026

Maximum number of one-element commutation classes of a permutation

In this paper, we provide an upper bound for the number of one-element commutation classes of a permutation, that is, the number of reduced words in which no commutation can be applied. Using this upper bound, we prove a conjecture that relates the number of reduced words with the number of commutation classes of a permutation.

math.CO↗

arXiv2017

Friendship and Selfishness Forwarding: applying machine learning techniques to Opportunistic Networks data forwarding

Opportunistic networks could become the solution to provide communication support in both cities where the cellular network could be overloaded, and in scenarios where a fixed infrastructure is not available, like in remote and developing regions. A critical issue that still requires a satisfactory solution is the design of an efficient data delivery solution. Social characteristics are recently being considered as a promising alternative. Most opportunistic network applications rely on the different mobile devices carried by users, and whose behavior affects the use of the device itself. This work presents the "Friendship and Selfishness Forwarding" (FSF) algorithm. FSF analyses two aspects to make message forwarding decisions when a contact opportunity arises: First, it classifies the friendship strength among a pair of nodes by using a machine learning algorithm to quantify the friendship strength among pairs of nodes in the network. Next, FSF assesses the relay node selfishness to consider those cases in which, despite a strong friendship with the destination, the relay node may not accept to receive the message because it is behaving selfishly, or because its device has resource constraints in that moment. By using trace-driven simulations through the ONE simulator, we show that the FSF algorithm outperforms previously proposed schemes in terms of delivery rate, average cost, and efficiency.

cs.NI↗