Biology subjects

Baptista, D.

Publications and source records attributed to Baptista, D..

3 recordsLinked to original sources

bioRxiv2025

AlphaFold models of host-pathogen interactions elucidate the prevalence and structural modes of molecular mimicry

Pathogens exploit host cellular machinery through protein-protein interactions (PPIs), often using molecular mimicry to hijack host cellular processes. While there have been thousands of host-pathogen PPIs determined to date, the lack of structural information for these impedes the study of the prevalence of molecular mimicry and convergent evolution of protein interaction interfaces. To address this, we benchmarked AlphaFold2 and 3 for prediction of structures of host-pathogen interactions observing that accurate models can be retrieved when ranking by modelling confidence, despite an overall low performance. We predicted structures for 6,782 pathogen-human PPIs yielding 803 models of higher confidence. Most pathogen proteins interacting with a common human protein are predicted to do so via the same interface, suggesting a high degree of convergent evolution of protein interaction interfaces. When comparing structural models from host-pathogen and host-host interactions, we observe that a majority of pathogen proteins are predicted to target existing human PPI interfaces. We categorized instances of mimicry into different modes, occurring at different frequencies: 1) via the same domain family (least common); 2) via a similar structural motif; and 3) via a similar linear motif (most common). We selected examples of linear motif interactions for binding assay testing, confirming 8 out of 12 predicted interfaces, including 3 viral linear motif interactions. This validates AlphaFolds ability to model some host-pathogen interactions and the mechanisms underlying molecular mimicry. This work showcases the value of large-scale structural modelling to study convergent evolution of host-pathogen interactions and how molecular mimicry may contribute to infection or host defense.

bioinformatics↗

bioRxiv2024

An atlas of protein-protein associations of human tissues prioritizes candidate disease genes

Proteins that interact together participate in the same cellular process and influence the same organismal traits. Despite the progress in mapping protein-protein interactions we lack knowledge of how they differ between tissues. Due to coordinated (post)transcriptional control, protein complex members have highly correlated abundances that are predictive of functional association. Here, we have compiled 7873 proteomic samples measuring protein levels in 11 human tissues and use these to define an atlas with tissue-specific protein associations. This method recapitulates known protein complexes and the larger structural organization of the cell. Interactions of stable protein complexes are well preserved across tissues, while signaling and metabolic interactions show larger variation. Further, we find that less than 18% of differences between tissues are estimated to be due to differences in gene expression while cell-type specific cellular structures, such as synaptic components, represent a significant driver of differences between tissues. We further supported the brain protein association network through co-fractionation experiments in synaptosomes, curation of brain derived pull-down data and AlphaFold2 models. Together these results illustrate how this brain specific protein interaction network can functionally prioritize candidate genes within loci linked to brain disorders.

systems biology↗

bioRxiv2022

A systematic evaluation of deep learning methods for the prediction of drug synergy in cancer

One of the main obstacles to the successful treatment of cancer is the phenomenon of drug resistance. A common strategy to overcome resistance is the use of combination therapies. However, the space of possibilities is huge and efficient search strategies are required. Machine Learning (ML) can be a useful tool for the discovery of novel, clinically relevant anti-cancer drug combinations. In particular, deep learning (DL) has become a popular choice for modeling drug combination effects. Here, we set out to examine the impact of different methodological choices on the performance of multimodal DL-based drug synergy prediction methods, including the use of different input data types, preprocessing steps and model architectures. Focusing on the NCI ALMANAC dataset, we found that feature selection based on prior biological knowledge has a positive impact on performance. Drug features appeared to be more predictive of drug response. Molecular fingerprint-based drug representations performed slightly better than learned representations, and gene expression data of cancer or drug response-specific genes also improved performance. In general, fully connected feature-encoding subnetworks outperformed other architectures, with DL outperforming other ML methods. Using a state-of-the-art interpretability method, we showed that DL models can learn to associate drug and cell line features with drug response in a biologically meaningful way. The strategies explored in this study will help to improve the development of computational methods for the rational design of effective drug combinations for cancer therapy. Author summaryCancer therapies often fail because tumor cells become resistant to treatment. One way to overcome resistance is by treating patients with a combination of two or more drugs. Some combinations may be more effective than when considering individual drug effects, a phenomenon called drug synergy. Computational drug synergy prediction methods can help to identify new, clinically relevant drug combinations. In this study, we developed several deep learning models for drug synergy prediction. We examined the effect of using different types of deep learning architectures, and different ways of representing drugs and cancer cell lines. We explored the use of biological prior knowledge to select relevant cell line features, and also tested data-driven feature reduction methods. We tested both precomputed drug features and deep learning methods that can directly learn features from raw representations of molecules. We also evaluated whether including genomic features, in addition to gene expression data, improves the predictive performance of the models. Through these experiments, we were able to identify strategies that will help guide the development of new deep learning models for drug synergy prediction in the future.

bioinformatics↗