We consider the conditional generation of 3D drug-like molecules with explicit control over molecular properties such as drug-like properties (e.g., Quantitative Estimate of Druglikeness or Synthetic Accessibility score) and effectively binding to specific protein sites. To tackle this problem, we propose an E(3)-equivariant Wasserstein autoencoder and factorize the latent space of our generative model into two disentangled aspects: molecular properties and the remaining structural context of 3D molecules. Our model ensures explicit control over these molecular attributes while maintaining equivariance of coordinate representation and invariance of data likelihood. Furthermore, we introduce a novel alignment-based coordinate loss to adapt equivariant networks for auto-regressive de-novo 3D molecule generation from scratch. Extensive experiments validate our model's effectiveness on property-guided and context-guided molecule generation, both for de-novo 3D molecule design and structure-based drug discovery against protein targets.

Designing protein-binding proteins with high affinity is critical in biomedical research and biotechnology. Despite recent advancements targeting specific proteins, the ability to create high-affinity binders for arbitrary protein targets on demand, without extensive rounds of wet-lab testing, remains a significant challenge. Here, we introduce PPDiff, a diffusion model to jointly design the sequence and structure of binders for arbitrary protein targets in a non-autoregressive manner. PPDiff builds upon our developed Sequence Structure Interleaving Network with Causal attention layers (SSINC), which integrates interleaved self-attention layers to capture global amino acid correlations, $k$-nearest neighbor ($k$NN) equivariant graph convolutional layers to model local interactions in three-dimensional (3D) space, and causal attention layers to simplify the intricate interdependencies within the protein sequence. To assess PPDiff, we curate PPBench, a general protein-protein complex dataset comprising 706,360 complexes from the Protein Data Bank (PDB). The model is pretrained on PPBench and finetuned on two real-world applications: target-protein mini-binder complex design and antigen-antibody complex design. PPDiff consistently surpasses baseline methods, achieving success rates of 50.00\%, 23.16\%, and 16.89\% for the pretraining task and the two downstream applications, respectively.

Artificial intelligence (AI)-driven methods can vastly improve the historically costly drug design process, with various generative models already in widespread use. Generative models for de novo drug design, in particular, focus on the creation of novel biological compounds entirely from scratch, representing a promising future direction. Rapid development in the field, combined with the inherent complexity of the drug design process, creates a difficult landscape for new researchers to enter. In this survey, we organize de novo drug design into two overarching themes: small molecule and protein generation. Within each theme, we identify a variety of subtasks and applications, highlighting important datasets, benchmarks, and model architectures and comparing the performance of top models. We take a broad approach to AI-driven drug design, allowing for both micro-level comparisons of various methods within each subtask and macro-level observations across different fields. We discuss parallel challenges and approaches between the two applications and highlight future directions for AI-driven de novo drug design as a whole. An organized repository of all covered sources is available at https://github.com/gersteinlab/GenAI4Drug.

In protein biophysics, the separation between the functionally important residues (forming the active site or binding surface) and those that create the overall structure (the fold) is a well-established and fundamental concept. Identifying and modifying those functional sites is critical for protein engineering but computationally non-trivial, and requires significant domain knowledge. To automate this process from a data-driven perspective, we propose a disentangled Wasserstein autoencoder with an auxiliary classifier, which isolates the function-related patterns from the rest with theoretical guarantees. This enables one-pass protein sequence editing and improves the understanding of the resulting sequences and editing actions involved. To demonstrate its effectiveness, we apply it to T-cell receptors (TCRs), a well-studied structure-function case. We show that our method can be used to alter the function of TCRs without changing the structural backbone, outperforming several competing methods in generation quality and efficiency, and requiring only 10\% of the running time needed by baseline models. To our knowledge, this is the first approach that utilizes disentangled representations for TCR engineering.

The COVID-19 pandemic caused by the SARS-CoV-2 virus has resulted in millions of deaths worldwide. The disease presents with various manifestations that can vary in severity and long-term outcomes. Previous efforts have contributed to the development of effective strategies for treatment and prevention by uncovering the mechanism of viral infection. We now know all the direct protein-protein interactions that occur during the lifecycle of SARS-CoV-2 infection, but it is critical to move beyond these known interactions to a comprehensive understanding of the "full interactome" of SARS-CoV-2 infection, which incorporates human microRNAs (miRNAs), additional human protein-coding genes, and exogenous microbes. Potentially, this will help in developing new drugs to treat COVID-19, differentiating the nuances of long COVID, and identifying histopathological signatures in SARS-CoV-2-infected organs. To construct the full interactome, we developed a statistical modeling approach called MLCrosstalk (multiple-layer crosstalk) based on latent Dirichlet allocation. MLCrosstalk integrates data from multiple sources, including microbes, human protein-coding genes, miRNAs, and human protein-protein interactions. It constructs "topics" that group SARS-CoV-2 with genes and microbes based on similar patterns of co-occurrence across patient samples. We use these topics to infer linkages between SARS-CoV-2 and protein-coding genes, miRNAs, and microbes. We then refine these initial linkages using network propagation to contextualize them within a larger framework of network and pathway structures. Using MLCrosstalk, we identified genes in the IL1-processing and VEGFA-VEGFR2 pathways that are linked to SARS-CoV-2. We also found that Rothia mucilaginosa and Prevotella melaninogenica are positively and negatively correlated with SARS-CoV-2 abundance, a finding corroborated by analysis of single-cell sequencing data.

MOTIVATION: Traditionally, an individual can only query and retrieve information from a genome browser by using accessories such as a mouse and keyboard. However, technology has changed the way that people interact with their screens. We hypothesized that we could leverage technological advances to use voice recognition as an interactive input to query and visualize genomic information. RESULTS: We developed an Amazon Alexa skill called Gene Tracer that allows users to use their voice to find disease-associated gene information, deleterious mutations and gene networks, while simultaneously enjoy a genome browser-like visualization experience on their screen. As the voice can be well recognized and understood, Gene Tracer provides users with more flexibility to acquire knowledge and is broadly applicable to other scenarios. AVAILABILITYAND IMPLEMENTATION: Alexa skill store (https://www.amazon.com/LT-Gene-tracer/dp/B08HCL1V68/) and a demonstration video (https://youtu.be/XbDbx7JDKmI). SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Recently, many chromatin immunoprecipitation sequencing experiments have been carried out for a diverse group of transcription factors (TFs) in many different types of human cells. These experiments manifest large-scale and dynamic changes in regulatory network connectivity (i.e. network 'rewiring'), highlighting the different regulatory programs operating in disparate cellular states. However, due to the dense and noisy nature of current regulatory networks, directly comparing the gains and losses of targets of key TFs across cell states is often not informative. Thus, here, we seek an abstracted, low-dimensional representation to understand the main features of network change. RESULTS: We propose a method called TopicNet that applies latent Dirichlet allocation to extract functional topics for a collection of genes regulated by a given TF. We then define a rewiring score to quantify regulatory-network changes in terms of the topic changes for this TF. Using this framework, we can pinpoint particular TFs that change greatly in network connectivity between different cellular states (such as observed in oncogenesis). Also, incorporating gene expression data, we define a topic activity score that measures the degree to which a given topic is active in a particular cellular state. And we show how activity differences can indicate differential survival in various cancers. AVAILABILITY AND IMPLEMENTATION: The TopicNet framework and related analysis were implemented using R and all codes are available at https://github.com/gersteinlab/topicnet. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: The pathogenesis of asthma is a complex process involving multiple genes and pathways. Identifying biomarkers from asthma datasets, especially those that include heterogeneous subpopulations, is challenging. Potentially, autoencoders provide ideal frameworks for such tasks as they can embed complex, noisy high-dimensional gene expression data into a low-dimensional latent space in an unsupervised fashion, enabling us to extract distinguishing features from expression data. RESULTS: Here, we developed a framework combining a denoising autoencoder and a supervised learning classifier to identify gene signatures related to asthma severity. Using the trained autoencoder with 50 hidden units, we found that hierarchical clustering on the low-dimensional embedding corresponds well with previously defined and clinically relevant clusters of patients. Moreover, each hidden unit has contributions from each of the genes, and pathway analysis of these contributions shows that the hidden units are significantly enriched in known asthma-related pathways. We then used genes that contribute most to the hidden units to develop a secondary random-forest classifier for directly predicting asthma severity. The feature importance metric from this classifier identified a signature based on 50 key genes, which are associated with severity. Furthermore, we can use these key genes to successfully estimate FEV1/FVC ratios across patients, via support-vector-machine regression. CONCLUSION: We found that the denoising autoencoder framework can extract meaningful patterns corresponding to functional gene groups and patient clusters from the gene expression of asthma patients.