Identification of piRNA Binding Sites Reveals the Argonaute Regulatory Landscape of the C. elegans Germline.

piRNAs (Piwi-interacting small RNAs) engage Piwi Argonautes to silence transposons and promote fertility in animal germlines. Genetic and computational studies have suggested that C. elegans piRNAs tolerate mismatched pairing and in principle could target every transcript. Here we employ in vivo cross-linking to identify transcriptome-wide interactions between piRNAs and target RNAs. We show that piRNAs engage all germline mRNAs and that piRNA binding follows microRNA-like pairing rules. Targeting correlates better with binding energy than with piRNA abundance, suggesting that piRNA concentration does not limit targeting. In mRNAs silenced by piRNAs, secondary small RNAs accumulate at the center and ends of piRNA binding sites. In germline-expressed mRNAs, however, targeting by the CSR-1 Argonaute correlates with reduced piRNA binding density and suppression of piRNA-associated secondary small RNAs. Our findings reveal physiologically important and nuanced regulation of individual piRNA targets and provide evidence for a comprehensive post-transcriptional regulatory step in germline gene expression.

The RNase PARN-1 Trims piRNA 3' Ends to Promote Transcriptome Surveillance in C. elegans.

Piwi-interacting RNAs (piRNAs) engage Piwi proteins to suppress transposons and are essential for fertility in diverse organisms. An interesting feature of piRNAs is that, while piRNA lengths are stereotypical within a species, they can differ widely between species. For example, piRNAs are mainly 29 and 30 nucleotides in humans, 24 to 30 nucleotides in D. melanogaster, and uniformly 21 nucleotides in C. elegans. However, how piRNA length is determined and whether length impacts function remains unknown. Here, we show that C. elegans deficient for PARN-1, a conserved RNase, accumulate untrimmed piRNAs with 3' extensions. Surprisingly, these longer piRNAs are stable and associate with the Piwi protein PRG-1 but fail to robustly recruit downstream silencing factors. Our findings identify PARN-1 as a key regulator of piRNA length in C. elegans and suggest that length is regulated to promote efficient transcriptome surveillance.

The HP1 homolog rhino anchors a nuclear complex that suppresses piRNA precursor splicing.

piRNAs guide an adaptive genome defense system that silences transposons during germline development. The Drosophila HP1 homolog Rhino is required for germline piRNA production. We show that Rhino binds specifically to the heterochromatic clusters that produce piRNA precursors, and that binding directly correlates with piRNA production. Rhino colocalizes to germline nuclear foci with Rai1/DXO-related protein Cuff and the DEAD box protein UAP56, which are also required for germline piRNA production. RNA sequencing indicates that most cluster transcripts are not spliced and that rhino, cuff, and uap56 mutations increase expression of spliced cluster transcripts over 100-fold. LacI::Rhino fusion protein binding suppresses splicing of a reporter transgene and is sufficient to trigger piRNA production from a trans combination of sense and antisense reporters. We therefore propose that Rhino anchors a nuclear complex that suppresses cluster transcript splicing and speculate that stalled splicing differentiates piRNA precursors from mRNAs.

KGDiff: towards explainable target-aware molecule generation with knowledge guidance.

Designing 3D molecules with high binding affinity for specific protein targets is crucial in drug design. One challenge is that the atomic interaction between molecules and proteins in 3D space has to be taken into account. However, the existing target-aware methods solely model the joint distribution between the molecules and proteins, disregarding the binding affinities between them, which leads to limited performance. In this paper, we propose an explainable diffusion model to generate molecules that can be bound to a given protein target with high affinity. Our method explicitly incorporates the chemical knowledge of protein-ligand binding affinity into the diffusion model, and uses the knowledge to guide the denoising process towards the direction of high binding affinity. Specifically, an SE(3)-invariant expert network is developed to fit the Vina scoring functions and jointly trained with the denoising network, while the domain knowledge is distilled and conveyed from Vina functions to the expert network. An effective guidance is proposed on both continuous atom coordinates and discrete atom types by taking advantages of the gradient of the expert network. Experiments on the benchmark CrossDocked2020 demonstrate the superiority of our method. Additionally, an atom-level explanation of the generated molecules is provided, and the connections with the domain knowledge are established.

Predicting cell line-specific synergistic drug combinations through a relational graph convolutional network with attention mechanism.

Identifying synergistic drug combinations (SDCs) is a great challenge due to the combinatorial complexity and the fact that SDC is cell line specific. The existing computational methods either did not consider the cell line specificity of SDC, or did not perform well by building model for each cell line independently. In this paper, we present a novel encoder-decoder network named SDCNet for predicting cell line-specific SDCs. SDCNet learns common patterns across different cell lines as well as cell line-specific features in one model for drug combinations. This is realized by considering the SDC graphs of different cell lines as a relational graph, and constructing a relational graph convolutional network (R-GCN) as the encoder to learn and fuse the deep representations of drugs for different cell lines. An attention mechanism is devised to integrate the drug features from different layers of the R-GCN according to their relative importance so that representation learning is further enhanced. The common patterns are exploited through partial parameter sharing in cell line-specific decoders, which not only reconstruct the known SDCs but also predict new ones for each cell line. Experiments on various datasets demonstrate that SDCNet is superior to state-of-the-art methods and is also robust when generalized to new cell lines that are different from the training ones. Finally, the case study again confirms the effectiveness of our method in predicting novel reliable cell line-specific SDCs.

MGAE-DC: Predicting the synergistic effects of drug combinations through multi-channel graph autoencoders.

Accurate prediction of synergistic effects of drug combinations can reduce the experimental costs for drug development and facilitate the discovery of novel efficacious combination therapies for clinical studies. The drug combinations with high synergy scores are regarded as synergistic ones, while those with moderate or low synergy scores are additive or antagonistic ones. The existing methods usually exploit the synergy data from the aspect of synergistic drug combinations, paying little attention to the additive or antagonistic ones. Also, they usually do not leverage the common patterns of drug combinations across different cell lines. In this paper, we propose a multi-channel graph autoencoder (MGAE)-based method for predicting the synergistic effects of drug combinations (DC), and shortly denote it as MGAE-DC. A MGAE model is built to learn the drug embeddings by considering not only synergistic combinations but also additive and antagonistic ones as three input channels. The later two channels guide the model to explicitly characterize the features of non-synergistic combinations through an encoder-decoder learning process, and thus the drug embeddings become more discriminative between synergistic and non-synergistic combinations. In addition, an attention mechanism is incorporated to fuse each cell-line's drug embeddings across various cell lines, and a common drug embedding is extracted to capture the invariant patterns by developing a set of cell-line shared decoders. The generalization performance of our model is further improved with the invariant patterns. With the cell-line specific and common drug embeddings, our method is extended to predict the synergy scores of drug combinations by a neural network module. Experiments on four benchmark datasets demonstrate that MGAE-DC consistently outperforms the state-of-the-art methods. In-depth literature survey is conducted to find that many drug combinations predicted by MGAE-DC are supported by previous experimental studies. The source code and data are available at https://github.com/yushenshashen/MGAE-DC.

piRTarBase: a database of piRNA targeting sites and their roles in gene regulation.

PIWI-interacting RNAs (piRNAs) are a class of small noncoding RNAs that guard animal genomes against mutation by silencing transposons. In addition, recent studies have reported that piRNAs silence various endogenous genes. Tens of thousands of distinct piRNAs made in animals do not pair well to transposons and currently the functions and targets of piRNAs are largely unexplored. piRTarBase provides a user-friendly interface to access both predicted and experimentally identified piRNA targeting sites in Caenorhabditis elegans. The user can input genes of interest and retrieve a list of piRNA targeting sites on the input genes. Alternatively, the user can input a piRNA and retrieve a list of its mRNA targets. Additionally, piRTarBase integrates published mRNA and small RNA sequencing data, which will help users identify biologically relevant targeting events. Importantly, our analyses suggest that the piRNA sites found by both predictive and experimental approaches are more likely to exhibit silencing effects on their targets than each method alone. Taken together, piRTarBase offers an integrative platform that will help users to identify functional piRNA target sites by evaluating various information. piRTarBase is freely available for academic use at http://cosbi6.ee.ncku.edu.tw/piRTarBase/.

pirScan: a webserver to predict piRNA targeting sites and to avoid transgene silencing in C. elegans.

pirScan is a web-based tool for identifying C. elegans piRNA-targeting sites within a given mRNA or spliced DNA sequence. The purpose of our tool is to allow C. elegans researchers to predict piRNA targeting sites and to avoid the persistent germline silencing of transgenes that has rendered many constructs unusable. pirScan fulfills this purpose by first enumerating the predicted piRNA-targeting sites present in an input sequence. This prediction can be exported in a tabular or graphical format. Subsequently, pirScan suggests silent mutations that can be introduced to the input sequence that would allow the modified transgene to avoid piRNA targeting. The user can customize the piRNA targeting stringency and the silent mutations that he/she wants to introduce into the sequence. The modified sequences can be re-submitted to be certain that any previously present piRNA-targeting sites are now absent and no new piRNA-targeting sites are accidentally generated. This revised sequence can finally be downloaded as a text file and/or visualized in a graphical format. pirScan is freely available for academic use at http://cosbi4.ee.ncku.edu.tw/pirScan/.

Glycolytic enzymes localize to ribonucleoprotein granules in Drosophila germ cells, bind Tudor and protect from transposable elements.

Germ cells give rise to all cell lineages in the next-generation and are responsible for the continuity of life. In a variety of organisms, germ cells and stem cells contain large ribonucleoprotein granules. Although these particles were discovered more than 100 years ago, their assembly and functions are not well understood. Here we report that glycolytic enzymes are components of these granules in Drosophila germ cells and both their mRNAs and the enzymes themselves are enriched in germ cells. We show that these enzymes are specifically required for germ cell development and that they protect their genomes from transposable elements, providing the first link between metabolism and transposon silencing. We further demonstrate that in the granules, glycolytic enzymes associate with the evolutionarily conserved Tudor protein. Our biochemical and single-particle EM structural analyses of purified Tudor show a flexible molecule and suggest a mechanism for the recruitment of glycolytic enzymes to the granules. Our data indicate that germ cells, similarly to stem cells and tumor cells, might prefer to produce energy through the glycolytic pathway, thus linking a particular metabolism to pluripotency.

Comparative functional characterization of the CSR-1 22G-RNA pathway in Caenorhabditis nematodes.

As a champion of small RNA research for two decades, Caenorhabditis elegans has revealed the essential Argonaute CSR-1 to play key nuclear roles in modulating chromatin, chromosome segregation and germline gene expression via 22G-small RNAs. Despite CSR-1 being preserved among diverse nematodes, the conservation and divergence in function of the targets of small RNA pathways remains poorly resolved. Here we apply comparative functional genomic analysis between C. elegans and Caenorhabditis briggsae to characterize the CSR-1 pathway, its targets and their evolution. C. briggsae CSR-1-associated small RNAs that we identified by immunoprecipitation-small RNA sequencing overlap with 22G-RNAs depleted in cbr-csr-1 RNAi-treated worms. By comparing 22G-RNAs and target genes between species, we defined a set of CSR-1 target genes with conserved germline expression, enrichment in operons and more slowly evolving coding sequences than other genes, along with a small group of evolutionarily labile targets. We demonstrate that the association of CSR-1 with chromatin is preserved, and show that depletion of cbr-csr-1 leads to chromosome segregation defects and embryonic lethality. This first comparative characterization of a small RNA pathway in Caenorhabditis establishes a conserved nuclear role for CSR-1 and highlights its key role in germline gene regulation across multiple animal species.

Efficient retrosynthetic planning with MCTS exploration enhanced A* search.

Retrosynthetic planning, which aims to identify synthetic pathways for target molecules from starting materials, is a fundamental problem in synthetic chemistry. Computer-aided retrosynthesis has made significant progress, in which heuristic search algorithms, including Monte Carlo Tree Search (MCTS) and A* search, have played a crucial role. However, unreliable guiding heuristics often cause search failure due to insufficient exploration. Conversely, excessive exploration also prevents the search from reaching the optimal solution. In this paper, MCTS exploration enhanced A* (MEEA*) search is proposed to incorporate the exploratory behavior of MCTS into A* by providing a look-ahead search. Path consistency is adopted as a regularization to improve the generalization performance of heuristics. Extensive experimental results on 10 molecule datasets demonstrate the effectiveness of MEEA*. Especially, on the widely used United States Patent and Trademark Office (USPTO) benchmark, MEEA* achieves a 100.0% success rate. Moreover, for natural products, MEEA* successfully identifies bio-retrosynthetic pathways for 97.68% test compounds.

Mammalian PIWI-piRNA-target complexes reveal features for broad and efficient target silencing.

The PIWI-interacting RNA (piRNA) pathway is an adaptive defense system wherein piRNAs guide PIWI family Argonaute proteins to recognize and silence ever-evolving selfish genetic elements and ensure genome integrity. Driven by this intensive host-pathogen arms race, the piRNA pathway and its targeted transposons have coevolved rapidly in a species-specific manner, but how the piRNA pathway adapts specifically to target silencing in mammals remains elusive. Here, we show that mouse MILI and human HILI piRNA-induced silencing complexes (piRISCs) bind and cleave targets more efficiently than their invertebrate counterparts from the sponge Ephydatia fluviatilis. The inherent functional differences comport with structural features identified by cryo-EM studies of piRISCs. In the absence of target, MILI and HILI piRISCs adopt a wider nucleic-acid-binding channel and display an extended prearranged piRNA seed as compared with EfPiwi piRISC, consistent with their ability to capture targets more efficiently than EfPiwi piRISC. In the presence of target, the seed gate-which enforces seed-target fidelity in microRNA RISC-adopts a relaxed state in mammalian piRISC, revealing how MILI and HILI tolerate seed-target mismatches to broaden the target spectrum. A vertebrate-specific lysine distorts the piRNA seed, shifting the trajectory of the piRNA-target duplex out of the central cleft and toward the PAZ lobe. Functional analyses reveal that this lysine promotes target binding and cleavage. Our study therefore provides a molecular basis for the piRNA targeting mechanism in mice and humans, and suggests that mammalian piRNA machinery can achieve broad target silencing using a limited supply of piRNA species.

IA-FaceS: A bidirectional method for semantic face editing.

Semantic face editing has achieved substantial progress in recent years. However, existing face editing methods, which often encode the entire image into a single code, still have difficulty in enabling flexible editing while keeping high-fidelity reconstruction. The one-code scheme also brings entangled face manipulations and limited flexibility in editing face components. In this paper, we present IA-FaceS, a bidirectional method for disentangled face attribute manipulation as well as flexible, controllable component editing. We propose to embed images onto two branches: one branch computes high-dimensional component-invariant content embedding for capturing face details, and the other provides low-dimensional component-specific embeddings for component manipulations. The two-branch scheme naturally enables high-quality facial component-level editing while keeping faithful reconstruction with details. Moreover, we devise a component adaptive modulation (CAM) module, which integrates component-specific guidance into the decoder and successfully disentangles highly-correlated face components. The single-eye editing is developed for the first time without editing face masks or sketches. According to the experimental results, IA-FaceS establishes a good balance between maintaining image details and performing flexible face manipulation. Both quantitative and qualitative results indicate that the proposed method outperforms the existing methods in reconstruction, face attribute manipulation, and component transfer. We release the code and weights at: https://github.com/CMACH508/IA-FaceS.

Self-Organizing a Latent Hierarchy of Sketch Patterns for Controllable Sketch Synthesis.

Encoding sketches as Gaussian mixture model (GMM)-distributed latent codes is an effective way to control sketch synthesis. Each Gaussian component represents a specific sketch pattern, and a code randomly sampled from the Gaussian can be decoded to synthesize a sketch with the target pattern. However, existing methods treat the Gaussians as individual clusters, which neglects the relationships between them. For example, the giraffe and horse sketches heading left are related to each other by their face orientation. The relationships between sketch patterns are important messages to reveal cognitive knowledge in sketch data. Thus, it is promising to learn accurate sketch representations by modeling the pattern relationships into a latent structure. In this article, we construct a tree-structured taxonomic hierarchy over the clusters of sketch codes. The clusters with the more specific descriptions of sketch patterns are placed at the lower levels, while the ones with the more general patterns are ranked at the higher levels. The clusters at the same rank relate to each other through the inheritance of features from common ancestors. We propose a hierarchical expectation-maximization (EM)-like algorithm to explicitly learn the hierarchy, jointly with the training of encoder-decoder network. Moreover, the learned latent hierarchy is utilized to regularize sketch codes with structural constraints. Experimental results show that our method significantly improves controllable synthesis performance and obtains effective sketch analogy results.

RefinePocket: An Attention-Enhanced and Mask-Guided Deep Learning Approach for Protein Binding Site Prediction.

Protein binding site prediction is an important prerequisite task of drug discovery and design. While binding sites are very small, irregular and varied in shape, making the prediction very challenging. Standard 3D U-Net has been adopted to predict binding sites but got stuck with unsatisfactory prediction results, incomplete, out-of-bounds, or even failed. The reason is that this scheme is less capable of extracting the chemical interactions of the entire region and hardly takes into account the difficulty of segmenting complex shapes. In this paper, we propose a refined U-Net architecture, called RefinePocket, consisting of an attention-enhanced encoder and a mask-guided decoder. During encoding, taking binding site proposal as input, we employ Dual Attention Block (DAB) hierarchically to capture rich global information, exploring residue relationship and chemical correlations in spatial and channel dimensions respectively. Then, based on the enhanced representation extracted by the encoder, we devise Refine Block (RB) in the decoder to enable self-guided refinement of uncertain regions gradually, resulting in more precise segmentation. Experiments show that DAB and RB complement and promote each other, making RefinePocket has an average improvement of 10.02% on DCC and 4.26% on DVO compared with the state-of-the-art method on four test sets.

DrugGen: a database of de novo-generated molecular binders for specified target proteins.

De novo molecular generation is a promising approach to drug discovery, building novel molecules from the scratch that can bind the target proteins specifically. With the increasing availability of machine learning algorithms and computational power, artificial intelligence (AI) has emerged as a valuable tool for this purpose. Here, we have developed a database of 3D ligands that collects six AI models for de novo molecular generation based on target proteins, including 20 disease-associated targets. Our database currently includes 1767 protein targets and up to 164 107 de novo-designed molecules. The primary goal is to provide an easily accessible and user-friendly molecular database for professionals in the fields of bioinformatics, pharmacology and related areas, enabling them to efficiently screen for potential lead compounds with biological activity. Additionally, our database provides a comprehensive resource for computational scientists to explore and compare different AI models in terms of their performance in generating novel molecules with desirable properties. All the resources and services are publicly accessible at https://cmach.sjtu.edu.cn/drug/. Database URL: https://cmach.sjtu.edu.cn/drug/.

A deep-learning method for the end-to-end prediction of intracranial aneurysm rupture risk.

It is critical to accurately predict the rupture risk of an intracranial aneurysm (IA) for timely and appropriate treatment because the fatality rate after rupture is 50 % . Existing methods relying on morphological features (e.g., height-width ratio) measured manually by neuroradiologists are labor intensive and have limited use for risk assessment. Therefore, we propose an end-to-end deep-learning method, called TransIAR net, to automatically learn the morphological features from 3D computed tomography angiography (CTA) data and accurately predict the status of IA rupture. We devise a multiscale 3D convolutional neural network (CNN) to extract the structural patterns of the IA and its neighborhood with a dual branch of shared network structures. Moreover, we learn the spatial dependence within the IA neighborhood with a transformer encoder. Our experiments demonstrated that the features learned by TransIAR are more effective and robust than handcrafted features, resulting in a 10 % - 15 % improvement in the accuracy of rupture status prediction.

Deep Rival Penalized Competitive Learning for low-resolution face recognition.

Current face recognition tasks are usually carried out on high-quality face images, but in reality, most face images are captured under unconstrained or poor conditions, e.g., by video surveillance. Existing methods are featured by learning data uncertainty to avoid overfitting the noise, or by adding margins to the angle or cosine space of the normalized softmax loss to penalize the target logit, which enforces intra-class compactness and inter-class discrepancy. In this paper, we propose a deep Rival Penalized Competitive Learning (RPCL) for deep face recognition in low-resolution (LR) images. Inspired by the idea of the RPCL, our method further enforces regulation on the rival logit, which is defined as the largest non-target logit for an input image. Different from existing methods that only consider penalization on the target logit, our method not only strengthens the learning towards the target label, but also enforces a reverse direction, i.e., becoming de-learning, away from the rival label. Comprehensive experiments demonstrate that our method improves the existing state-of-the-art methods to be very robust for LR face recognition.

AlphaDrug: protein target specific de novo molecular generation.

Traditional drug discovery is very laborious, expensive, and time-consuming, due to the huge combinatorial complexity of the discrete molecular search space. Researchers have turned to machine learning methods for help to tackle this difficult problem. However, most existing methods are either virtual screening on the available database of compounds by protein-ligand affinity prediction, or unconditional molecular generation, which does not take into account the information of the protein target. In this paper, we propose a protein target-oriented de novo drug design method, called AlphaDrug. Our method is able to automatically generate molecular drug candidates in an autoregressive way, and the drug candidates can dock into the given target protein well. To fulfill this goal, we devise a modified transformer network for the joint embedding of protein target and the molecule, and a Monte Carlo tree search (MCTS) algorithm for the conditional molecular generation. In the transformer variant, we impose a hierarchy of skip connections from protein encoder to molecule decoder for efficient feature transfer. The transformer variant computes the probabilities of next atoms based on the protein target and the molecule intermediate. We use the probabilities to guide the look-ahead search by MCTS to enhance or correct the next-atom selection. Moreover, MCTS is also guided by a value function implemented by a docking program, such that the paths with many low docking values are seldom chosen. Experiments on diverse protein targets demonstrate the effectiveness of our methods, indicating that AlphaDrug is a potentially promising solution to target-specific de novo drug design.

GLH/VASA helicases promote germ granule formation to ensure the fidelity of piRNA-mediated transcriptome surveillance.

piRNAs function as guardians of the genome by silencing non-self nucleic acids and transposable elements in animals. Many piRNA factors are enriched in perinuclear germ granules, but whether their localization is required for piRNA biogenesis or function is not known. Here we show that GLH/VASA helicase mutants exhibit defects in forming perinuclear condensates containing PIWI and other small RNA cofactors. These mutant animals produce largely normal levels of piRNA but are defective in triggering piRNA silencing. Strikingly, while many piRNA targets are activated in GLH mutants, we observe that hundreds of endogenous genes are aberrantly silenced by piRNAs. This defect in self versus non-self recognition is also observed in other mutants where perinuclear germ granules are disrupted. Together, our results argue that perinuclear germ granules function critically to promote the fidelity of piRNA-based transcriptome surveillance in C. elegans and preserve self versus non-self distinction.