Predicting transcriptional activation domain function using Graph Neural Networks.

Analysis of factors that lead to the functionality of transcriptional activation domains remains a crucial and yet challenging task owing to the significant diversity in their sequences and their intrinsically disordered nature. Almost all existing methods that have aimed to predict activation domains have involved traditional machine learning approaches, such as logistic regression, that are unable to capture complex patterns in data or plain convolutional neural networks and have been limited in exploration of structural features. However, there is a tremendous potential in the inspection of the structural properties of activation domains, and an opportunity to investigate complex relationships between features of residues in the sequence. To address these, we have utilized the power of graph neural networks which can represent structural data in the form of nodes and edges, allowing nodes to exchange information among themselves. We have experimented with two kinds of graph formulations, one involving residues as nodes and the other assigning atoms to be the nodes. A logistic regression model was also developed to analyze feature importance. For all the models, several feature combinations were experimented with. The residue-level GNN model with amino acid type, residue position, acidic/basic/aromatic property and secondary structure feature combination gave the best performing model with accuracy, F1 score and AUROC of 97.9%, 71% and 97.1% respectively which outperformed other existing methods in the literature when applied on the dataset we used. Among the other structure-based features that were analyzed, the amphipathic property of helices also proved to be an important feature for classification. Logistic regression results showed that the most dominant feature that makes a sequence functional is the frequency of different types of amino acids in the sequence. Our results consistent have shown that functional sequences have more acidic and aromatic residues whereas basic residues are seen more in non-functional sequences.

Diversity and selection analyses identify transmission-blocking antigens as the optimal vaccine candidates in Plasmodium falciparum.

Background: A highly effective vaccine for malaria remains an elusive target, at least in part due to the under-appreciated natural parasite variation. This study aimed to investigate genetic and structural variation, and immune selection of leading malaria vaccine candidates across the Plasmodium falciparum's life cycle. Methods: We analyzed 325 P. falciparum whole genome sequences from Zambia, in addition to 791 genomes from five other African countries available in the MalariaGEN Pf3k Rdatabase. Ten vaccine antigens spanning three life-history stages were examined for genetic and structural variations, using population genetics measures, haplotype network analysis, and 3D structure selection analysis. Findings: Among the ten antigens analyzed, only three in the transmission-blocking vaccine category display P. falciparum 3D7 as the dominant haplotype. The antigens AMA1, CSP, MSP119 and CelTOS, are much more diverse than the other antigens, and their epitope regions are under moderate to strong balancing selection. In contrast, Rh5, a blood stage antigen, displays low diversity yet slightly stronger immune selection in the merozoite-blocking epitope region. Except for CelTOS, the transmission-blocking antigens Pfs25, Pfs48/45, Pfs230, Pfs47, and Pfs28 exhibit minimal diversity and no immune selection in epitopes that induce strain-transcending antibodies, suggesting potential effectiveness of 3D7-based vaccines in blocking transmission. Interpretations: These findings offer valuable insights into the selection of optimal vaccine candidates against P. falciparum. Based on our results, we recommend prioritizing conserved merozoite antigens and transmission-blocking antigens. Combining these antigens in multi-stage approaches may be particularly promising for malaria vaccine development initiatives. Funding: Purdue Department of Biological Sciences; Puskas Memorial Fellowship; National Institute of Allergy and Infectious Diseases (U19AI089680).

Activation of gene expression by detergent-like protein domains.

The mechanisms by which transcriptional activation domains (tADs) initiate eukaryotic gene expression have been an enigma for decades because most tADs lack specificity in sequence, structure, and interactions with targets. Machine learning analysis of data sets of tAD sequences generated in vivo elucidated several functionality rules: the functional tAD sequences should (i) be devoid of or depleted with basic amino acid residues, (ii) be enriched with aromatic and acidic residues, (iii) be with aromatic residues localized mostly near the terminus of the sequence, and acidic residues localized more internally within a span of 20-30 amino acids, (iv) be with both aromatic and acidic residues preferably spread out in the sequence and not clustered, and (v) not be separated by occasional basic residues. These and other more subtle rules are not absolute, reflecting absence of a tAD consensus sequence, enormous variability, and consistent with surfactant-like tAD biochemical properties. The findings are compatible with the paradigm-shifting nucleosome detergent mechanism of gene expression activation, contributing to the development of the liquid-liquid phase separation model and the biochemistry of near-stochastic functional allosteric interactions.