Biology subjects

Peng, J.

Publications and source records attributed to Peng, J..

At least 127 records · Page 7Linked to original sources

bioRxiv2023

CDCA7 is a hemimethylated DNA adaptor for the nucleosome remodeler HELLS

Mutations of the SNF2 family ATPase HELLS and its activator CDCA7 cause immunodeficiency-centromeric instability-facial anomalies (ICF) syndrome, characterized by hypomethylation at heterochromatin. The unique zinc-finger domain, zf-4CXXC_R1, of CDCA7 is widely conserved across eukaryotes but is absent from species that lack HELLS and DNA methyltransferases, implying its specialized relation with methylated DNA. Here we demonstrate that zf-4CXXC_R1 acts as a hemimethylated DNA sensor. The zf-4CXXC_R1 domain of CDCA7 selectively binds to DNA with a hemimethylated CpG, but not unmethylated or fully methylated CpG, and ICF disease mutations eliminated this binding. CDCA7 and HELLS interact via their N-terminal alpha helices, through which HELLS is recruited to hemimethylated DNA. While placement of a hemimethylated CpG within the nucleosome core particle can hinder its recognition by CDCA7, cryo-EM structure analysis of the CDCA7-nucleosome complex suggests that zf-4CXXC_R1 recognizes a hemimethylated CpG in the major groove at linker DNA. Our study provides insights into how the CDCA7-HELLS nucleosome remodeling complex uniquely assists maintenance DNA methylation.

molecular biology↗

bioRxiv2023

The ubiquitin-conjugating enzyme UBE2D/eff maintains a youthful proteome and ensures protein quality control during aging

Ubiquitin-conjugating enzymes (E2s) are key for regulating protein function and turnover via ubiquitination but it remains undetermined which E2s maintain proteostasis during aging. Here, we find that E2s have diverse roles in handling a model aggregation-prone protein (huntingtin-polyQ) in the Drosophila retina: while some E2s mediate aggregate assembly, UBE2D/effete (eff) and other E2s are required for huntingtin-polyQ degradation. UBE2D/eff is key for proteostasis also in skeletal muscle: eff protein levels decline with aging, and muscle-specific eff knockdown causes an accelerated buildup in insoluble poly-ubiquitinated proteins (which progressively accumulate with aging) and shortens lifespan. Transgenic expression of human UBE2D2, homologous to eff, partially rescues the lifespan and proteostasis deficits caused by muscle-specific effRNAi by re-establishing the physiological levels of effRNAi-regulated proteins, which include several regulators of proteostasis. Interestingly, UBE2D/eff knockdown in young age reproduces part of the proteomic changes that normally occur in old muscles, suggesting that the decrease in UBE2D/eff protein levels that occurs with aging contributes to reshaping the composition of the muscle proteome. Altogether, these findings indicate that UBE2D/eff is a key E2 ubiquitin-conjugating enzyme that ensures protein quality control and helps maintain a youthful proteome composition during aging.

cell biology↗

bioRxiv2023

The Origin and Evolution of Sex Peptide and Sex Peptide Receptor Interactions

Post-mating responses play a vital role in successful reproduction across diverse species. In fruit flies, sex peptide (SP) binds to the sex peptide receptor (SPR), triggering a series of post-mating responses. However, the origin of SPR predates the emergence of SP. The evolutionary origins of the interactions between SP and SPR and the mechanisms by which they interact remain enigmatic. In this study, we used ancestral sequence reconstruction, AlphaFold2 predictions, and molecular dynamics simulations to study SP-SPR interactions and their origination. Using AlphaFold2 and long-time molecular dynamics (MD) simulations, we predicted the structure and dynamics of SP-SPR interactions. We show that SP potentially binds to the ancestral states of Diptera SPR. Notably, we found that only a few amino acid changes in SPR are sufficient for the formation of SP-SPR interactions. Ancestral sequence reconstruction and MD simulations further reveal that SPR interacts with SP through residues that are mostly involved in the interaction interface of an ancestral ligand, myoinhibitory peptides (MIPs). We propose a potential mechanism whereby SP-SPR interactions arise from the pre-existing MIP-SPR interface as well as early chance events both inside and outside the pre-existing interface that created novel SP-specific SP-SPR interactions. Our findings provide new insights into the origin and evolution of SP-SPR interactions and their relationship with MIP-SPR interactions.

evolutionary biology↗

bioRxiv2023

Platelet-specific TGFβ1 deficiency aggravates atherosclerosis, vascular inflammation, and hypercholesterolemia in mice

Atherosclerosis involves inflammatory and thrombotic mechanisms, to which both platelets and transforming growth factor {beta} (TGF{beta}) contribute. The effect of platelet-derived TGF{beta} on atherosclerosis is, however, unknown and therefore investigated. Murine platelet-selective TGF{beta}-deficiency (plt-TGF{beta}-/-) was created by a Pf4-Cre approach, and an atherosclerotic mouse model was established by functional abrogation of Ldlr and 10-15 weeks of a high-fat diet in plt-TGF{beta}-/- mice and their non-plt-TGF{beta}-/- littermates. En face Oil Red O staining of the aorta showed more atherosclerotic lesion formation in plt-TGF{beta}-/- mice, with significant increases in both lesion size and lesion coverage of the total aortic area. Cryosections of the aortic root confirmed the aggravation of atherogenesis. Platelet-derived TGF{beta} deficiency increased circulating platelets and plasma levels of total cholesterol, LDL-cholesterol, and triglycerides after a 10 or 15 week high-fat diet period. RNA sequencing and proteomic analyses of the aorta showed signs of CD4+ T effector cell and macrophage activation in plt-TGF{beta}-/- mice. In conclusion, platelet-specific TGF{beta} deficiency aggravates atherosclerosis, via increasing arterial inflammation and plasma levels of cholesterol. Our findings demonstrate that platelet-derived TGF{beta} is prominently athero-protective. Key pointsO_LIPlatelet-specific transforming growth factor {beta} (TGF{beta}) deficiency markedly enhances atherosclerosis in a high-fat diet-fed murine model. C_LIO_LIPlatelet TGF{beta} deficiency aggravates hyperlipidemia, with further elevations of total cholesterol, LDL-cholesterol, and triglycerides. C_LI

cell biology↗

bioRxiv2023

Multimodal joint deconvolution and integrative signature selection in proteomics

Deconvolution is an efficient approach for detecting cell-type-specific (cs) transcriptomic signals without cellular segmentation. However, this type of methods have not been extended to the proteomics research. Here we present a novel algorithm and tool to dissect bulk proteome by leveraging the information shared between transcriptome-proteome. Our tool first identifies potential cell marker proteins by integrating RNA and protein bulk expression profiles and then jointly quantifies the cell abundance in mixture proteomes without using a reference signature matrix, enabling the downstream analyses such as cs-protein Quantitative Trait Loci (cspQTL) mapping. This new method and the cspQTL analysis are implemented in the R package MIC-SQTL that also provides integrative visualization of bulk multimodal samples, available at https://bioconductor.org/packages/MICSQTL.

bioinformatics↗

bioRxiv2023

Single-cell bisulfite-free 5mC and 5hmC sequencing with high sensitivity and scalability

Existing single cell bisulfite-based DNA methylation analysis is limited by low DNA recovery, and the measurement of 5hmC at single-base resolution remains challenging. Here we present a bisulfite-free single-cell whole-genome 5mC and 5hmC profiling technique, named Cabernet, which can characterize 5mC and 5hmC at single-base resolution with high genomic coverage. Cabernet utilizes Tn5 transposome for DNA fragmentation, which enables the discrimination between different alleles for measuring hemi-methylation status. Using Cabernet, we revealed the 5mC, hemi-5mC and 5hmC dynamics during early mouse embryo development, uncovering genomic regions exclusively governed by active or passive demethylation. We show that hemi-methylation status can be used to distinguish between pre- and post-replication cells, enabling more efficient cell grouping when integrated with 5mC profile. The property of Tn5 naturally enables Cabernet to achieve high-throughput single-cell methylome profiling, where we probed mouse cortical neurons and embryonic day 7.5 (E7.5) embryos, and constructed the library for thousands of single cells at high efficiency, demonstrating its potential for analyzing complex tissues at substantially low cost. Together, we present a new way of high-throughput methylome and hydroxymethylome detection at single-cell resolution, enabling efficient analysis of epigenetic status of biological systems with complicated nature such as neurons and cancer cells. Significance StatementMost of current methylation profiling techniques rely on bisulfite treatment, which suffers low DNA recovery. The technique proposed in this study, named Cabernet, can be used to measure 5mC and 5hmC at single-base resolution with high genomic coverage. By using Tn5 transposome, hemi-methylation status can be measured and high-throughput methylome profiling can be achieved. Together, it provides an efficient way to analyze the epigenetic landscape of complicated biological systems.

genomics↗

bioRxiv2023

Molecular de-extinction of antibiotics enabled by deep learning

Molecular de-extinction is an emerging field that aims to resurrect molecules to solve present-day problems such as antibiotic resistance. Here, we introduce a deep learning approach called Antibiotic Peptide de-Extinction (APEX) to mine the proteomes of all available extinct organisms (the "extinctome") searching for encrypted peptide (EP) antibiotics. APEX mined a total of 10,311,899 EPs and identified 37,176 sequences predicted to have broad-spectrum antimicrobial activity, 11,035 of which were not found in extant organisms. Chemical synthesis and experimental validation yielded archaic EPs (AEPs) with activity against dangerous bacterial pathogens. Most peptides killed bacteria by depolarizing their cytoplasmic membrane, contrary to known antimicrobial peptides, which target the outer membrane. Notably, lead peptides, including those derived from the woolly mammoth, ancient sea cow, giant sloth, and extinct giant elk, exhibited anti-infective activity in preclinical mouse models. We propose molecular de-extinction, accelerated by deep learning, as a framework for discovering therapeutic molecules.

bioengineering↗

bioRxiv2023

Pervasive mislocalization of pathogenic coding variants underlying human disorders

Widespread sequencing has yielded thousands of missense variants predicted or confirmed as disease-causing. This creates a new bottleneck: determining the functional impact of each variant - largely a painstaking, customized process undertaken one or a few genes or variants at a time. Here, we established a high-throughput imaging platform to assay the impact of coding variation on protein localization, evaluating 3,547 missense variants of over 1,000 genes and phenotypes. We discovered that mislocalization is a common consequence of coding variation, affecting about one-sixth of all pathogenic missense variants, all cellular compartments, and recessive and dominant disorders alike. Mislocalization is primarily driven by effects on protein stability and membrane insertion rather than disruptions of trafficking signals or specific interactions. Furthermore, mislocalization patterns help explain pleiotropy and disease severity and provide insights on variants of unknown significance. Our publicly available resource will likely accelerate the understanding of coding variation in human diseases.

cell biology↗

bioRxiv2023

The balance between shear flow and extracellular matrix in ovarian cancer-on-chip

Ovarian cancer is the most lethal gynecologic cancer in developed countries. Silent onset of the metastatic activity of ovarian tumor cells is factor for poor outcomes. In the tumor microenvironment, the extracellular matrix (ECM) and flow shear stress are known to play key roles in directing cell invasion. Hence, acute and tunable tools are critical to mimic scaffold and fluid for building clinically relevant in vitro models. We have built an ovarian tumor-on-chip where tunable ECM models are easily seeded with tumor spheroids and integrated within a microfluidic chip. This allows the investigation of the crosstalk between the characteristics of the ECM models and shear stress on the migratory behavior and cellular heterogeneity of ovarian tumor cells. We vary the composition of the ECM playing with type I and IV collagens and laminin, and control the shear stress in the chip. This work shows that in the shear stress regime of the peritoneal cavity, the ECM plays a major role in driving individual or collective modes of migration. In the presence of basement membrane proteins, migration is more collective that on type I collagen regardless of shear stress level. In addition, with increasing shear stress, individual cell migration was enhanced, while no significant impact on collective migration could be measured. This highlights our ability to discriminate relevant parameters for onset and shifts of cell behavior using our in vitro models. Furthermore, we described the ability to shift cells from an epithelial phenotype to a more mesenchymal phenotype, which could allow us to describe the role of these parameters during epithelial-to-mesenchymal (EMT) transition as a continuous process. Finally, we conclude that the ECM should hold a central position in in vitro cancer models, to understand cell response and develop platforms for therapeutic development.

bioengineering↗

bioRxiv2023

Metabolic reprogramming of cancer cells by JMJD6-mediated pre-mRNA splicing is associated with therapeutic response to splicing inhibitor

Dysregulated pre-mRNA splicing and metabolism are two hallmarks of MYC-driven cancers. Pharmacological inhibition of both processes has been extensively investigated as potential therapeutic avenues in preclinical and clinical studies. However, how pre-mRNA splicing and metabolism are orchestrated in response to oncogenic stress and therapies is poorly understood. Here, we demonstrate that Jumonji Domain Containing 6, Arginine Demethylase and Lysine Hydroxylase, JMJD6, acts as a hub connecting splicing and metabolism in MYC-driven neuroblastoma. JMJD6 cooperates with MYC in cellular transformation by physically interacting with RNA binding proteins involved in pre-mRNA splicing and protein homeostasis. Notably, JMJD6 controls the alternative splicing of two isoforms of glutaminase (GLS), namely kidney-type glutaminase (KGA) and glutaminase C (GAC), which are rate-limiting enzymes of glutaminolysis in the central carbon metabolism in neuroblastoma. Further, we show that JMJD6 is correlated with the anti-cancer activity of indisulam, a "molecular glue" that degrades splicing factor RBM39, which complexes with JMJD6. The indisulam-mediated cancer cell killing is at least partly dependent on the glutamine-related metabolic pathway mediated by JMJD6. Our findings reveal a cancer-promoting metabolic program is associated with alternative pre-mRNA splicing through JMJD6, providing a rationale to target JMJD6 as a therapeutic avenue for treating MYC-driven cancers.

cancer biology↗

bioRxiv2023

12R-HETE acts as an endogenous ligand of Nur77 in intestine and regulates ILC3s plasticity

Group 3 innate lymphoid cells (ILC3s), a heterogeneous population, are tissue-resident myeloid cells and have an essential role in bacterial infection. Although the plasticity of NKp46- CCR6- double-negative (DN) ILC3s toward the NKp46+ ILC3s is an important process in the development of intestinal immunity, the underlying molecular mechanisms responsible for this process remain poorly understood. Nur77 is an orphan receptor which regulates intestinal ILC3s expansion. However, the impact of Nur77 on the plasticity of intestinal ILC3s remains unclear. Here, we generated Nur77 null mice and investigated ILC3s expansion. The deficiency of Nur77 inhibited the mouse small intestinal ILC3s expansion and conversion of NKp46- ILC3s to NKp46+ ILC3s. We identified that 12R-HETE derived from arachidonic acid (ARA) in mouse intestine is an endogenous ligand of Nur77 and activates its transcriptional activity. The treatment with 12R-HETE promoted the differentiation of NKp46- ILC3s into NKp46+ ILC3s by enhancing the T-bet expression, thereby increased IFN-{gamma} production from NKp46+ ILC3s, and reduced the susceptibility to bacterial infection in WT, but not Nur77-/-, suckling mice. An integrated analysis of ATAC-seq and Smart RNA-seq showed that Rflnb, Impdh1, Map1s, and Gtpbp3 might be downstream targeted genes of Nur77 in response to 12R-HETE and mediate the regulation of ILC3s plasticity. In the presence of mycophenolic acid, an inhibitor of IMPDH, 12R-HETE no longer regulated the percentages of ROR{gamma}t+ILC3s and NKp46+ILC3s. We conclude that 12R-HETE acts as an endogenous ligand of Nur77, and regulates the ILC3s expansion and plasticity, and in turn, gut homeostasis and pathogen defense.

immunology↗

bioRxiv2023

A Likelihood Ratio Test for Hybridization Under the Multispecies Coalescent

Several methods have been developed to carry out a statistical test for hybridization at the species level, including the ABBA-BABA test and HyDe. Here, we propose a new method for detecting hybridization and quantifying the extent of hybridization. Our test computes the likelihood of a species tree that is possibly subject to hybridization using site pattern frequencies from genomic-scale datasets under the multispecies coalescent. To do this, we extend the calculation of the likelihood for site pattern frequency data for the 4-taxon symmetric and asymmetric species trees proposed in Chifman and Kubatko (2015) by incorporating an inheritance parameter, resulting in efficient computation of the likelihood under a scenario of hybridization. We use this likelihood computation to construct a likelihood ratio test that a given species is a hybrid of two parental species. Simulations demonstrate that our test is more powerful than existing tests of hybridization, including HyDe, and that it achieves the desired type I error rate. We apply the method to two empirical data sets, one for which hybridization is believed to have occurred and one for which previous methods have failed to detect hybridization.

evolutionary biology↗

bioRxiv2023

The impact of common variants on gene expression in the human brain: from RNA to protein to schizophrenia risk

BackgroundThe impact of genetic variants on gene expression has been intensely studied at the transcription level, yielding invaluable insights into the association between genes and the risk of complex disorders, such as schizophrenia (SCZ). However, the downstream impact of these variants and the molecular mechanisms connecting transcription variation to disease risk are not well understood. ResultsWe quantitated ribosome occupancy in prefrontal cortex samples of the BrainGVEX cohort. Together with transcriptomics and proteomics data from the same cohort, we performed cis-Quantitative Trait Locus (QTL) mapping and identified, at 10% false discovery rate, 3,253 expression QTLs (eQTLs), 1,344 ribosome occupancy QTLs (rQTLs), and 657 protein QTLs (pQTLs) out of 7,458 genes from 185 samples. Of the eQTLs identified, only 34% have their effects propagated to the protein level. Further analysis on the effect size of prefrontal cortex eQTLs identified from an independent dataset clearly replicated the post-transcriptional attenuation of eQTL effects. We identified omics-specific QTLs and investigated their potential in driving disease risks. Using both a variant-based approach and a gene-based approach, we identified genes containing expression-specific QTLs (esQTLs), ribosome-occupancy-specific QTLs (rsQTLs), and protein-specific QTLs (psQTLs). Among the variant-based omics-specific QTL, 38 showed strong colocalization with brain associated disorder GWAS signals, 29 of them are esQTLs. From the gene-based approach, we found 11 brain associated disorder risk genes that are driven predominantly by omics-specific QTL, all of them are driven by variants impacting transcriptional regulation. To take a complementary approach to further investigate the functional relevance of genes driven predominantly by attenuated eQTL signals, we identified SCZ risk genes using each omics independently and then investigated the omics-specificity of the driver regulatory process for each risk gene. Using S-PrediXcan we identified 74 SCZ risk genes across the three omics, 30% of which were novel, and 67% of these risk genes were confirmed to be causal in a MR-Egger test. Notably, 52 out of the 74 risk genes were identified using eQTL data and 68% of these SCZ-risk-gene-driving eQTLs show little to no evidence of driving corresponding variations at the protein level. ConclusionThe effect of eQTLs on gene expression in the prefrontal cortex is commonly attenuated post-transcriptionally. Many of the attenuated eQTLs still correlate with GWAS signals of brain associated complex disorders, indicating the possibility that these eQTL variants drive disease risk through mechanisms other than regulating protein expression level. Further investigation is needed to elucidate the mechanistic link between attenuated eQTLs and brain associated complex disorders.

genetics↗

bioRxiv2023

Saturating the eQTL map in Drosophila melanogaster: genome-wide patterns of cis and trans regulation of transcriptional variation in outbred populations

Decades of genome-wide mapping have shown that most genetic polymorphisms associated with complex traits are found in non-coding regions of the genome. Characterizing the effect of such genetic variation presents a formidable challenge, and eQTL mapping has been a key approach to understand the non-coding genome. However, comprehensive eQTL maps are available only for a few species like yeast and humans. With the aim of understanding the genetic landscape that regulates transcriptional variation in Drosophila melanogaster, we developed an outbred mapping panel in this species, the Drosophila Outbred Synthetic Panel (Dros-OSP). Using this community resource, we collected transcriptomic and genomic data for 1800 individual flies and were able to map cis and trans eQTLs for 98% of the genes expressed in D. melanogaster, increasing by thousands the number of genes for which regulatory loci are known in this species. We described, for the first time in the context of an outbred population, the properties of local and distal regulation of gene expression in terms of genetic diversity, heritability, connectivity, and pleiotropy. We uncovered that, contrary to long-standing assumptions, a significant part of gene co-expression networks is organized in a non-modular fashion. These results bring the fruit fly to the level of understanding that was only available for a few other organisms, and offer a new mapping resource that will expand the possibilities currently available to the Drosophila community. This data is available at Drosophilaeqtl.org.

genomics↗

bioRxiv2023

Phosphorylation of 53BP1 by ATM enforce neurodevelopmental programs in cortical organoids

53BP1 is a well-established DNA damage repair factor recently shown to regulate gene expression and critically influence tumor suppression and neural development. For gene regulation, how 53BP1 is regulated remains unclear. Here, we showed that 53BP1-serine 25 phosphorylation by ATM is required for neural progenitor cell proliferation and neuronal differentiation in cortical organoids. 53BP1-serine 25 phosphorylation dynamics controls 53BP1 target genes for neuronal differentiation and function, cellular response to stress, and apoptosis. Beyond 53BP1, ATM is required for phosphorylation of factors in neuronal differentiation, cytoskeleton, p53 regulation, and ATM, BNDF, and WNT signaling pathways for cortical organoid differentiation. Overall, our data suggest that 53BP1 and ATM control key genetic programs required for human cortical development.

developmental biology↗

bioRxiv2023

The origin and structural evolution of de novo genes in Drosophila

Although previously thought to be unlikely, recent studies have shown that de novo gene origination from previously non-genic sequences is a relatively common mechanism for gene innovation in many species and taxa. These young genes provide a unique set of candidates to study the structural and functional origination of proteins. However, our understanding of their protein structures and how these structures originate and evolve are still limited, due to a lack of systematic studies. Here, we combined high-quality base-level whole genome alignments, bioinformatic analysis, and computational structure modeling to study the origination, evolution, and protein structure of lineage-specific de novo genes. We identified 555 de novo gene candidates in D. melanogaster that originated within the Drosophilinae lineage. We found a gradual shift in sequence composition, evolutionary rates, and expression patterns with their gene ages, which indicates possible gradual shifts or adaptations of their functions. Surprisingly, we found little overall protein structural changes for de novo genes in the Drosophilinae lineage. Using Alphafold2, ESMFold, and molecular dynamics, we identified a number of de novo gene candidates with protein products that are potentially well-folded, many of which are more likely to contain transmembrane and signal proteins compared to other annotated protein-coding genes. Using ancestral sequence reconstruction, we found that most potentially well-folded proteins are often born folded. Interestingly, we observed one case where disordered ancestral proteins become ordered within a relatively short evolutionary time. Single-cell RNA-seq analysis in testis showed that although most de novo genes are enriched in spermatocytes, several young de novo genes are biased in the early spermatogenesis stage, indicating potentially important but less emphasized roles of early germline cells in the de novo gene origination in testis. This study provides a systematic overview of the origin, evolution, and structural changes of Drosophilinae-specific de novo genes.

evolutionary biology↗

bioRxiv2023

The path to "femmes fatales": the evolution of toxin resistance in predatory fireflies.

Toxic cardiotonic steroids (CTS) act as a defense mechanism in many firefly species (Lampyridae) by inhibiting a crucial enzyme called Na+,K+-ATPase (NKA). While most fireflies produce these toxins internally, species of the genus Photuris acquire them from a surprising source: predation on other fireflies. The contrasting physiology of toxin exposure and sequestration between Photuris and other firefly genera suggests that distinct strategies may be required to prevent self-intoxication. Our study demonstrates that both Photuris and their firefly prey have evolved highly-resistant NKAs. Using an evolutionary analysis of the specific target of CTS (ATP) in fireflies, and gene-editing in Drosophila, we find that the initial steps towards resistance were shared among Photuris and other firefly lineages. However, the Photuris lineage subsequently underwent multiple rounds of gene duplication and neofunctionalization, resulting in the development of ATP paralogs that are differentially expressed and exhibit increasing resistance to CTS. In contrast, other firefly species have maintained a single copy. Our results implicate gene duplication as a facilitator in the transition of Photuris to its distinct ecological role as predator of toxic firefly prey. One-Sentence SummaryGene duplication and neofunctionalization distinguish firefly predators from their toxic firefly prey.

evolutionary biology↗

bioRxiv2023

Rotamer Density Estimator is an Unsupervised Learner of the Effect of Mutations on Protein-Protein Interaction

AO_SCPLOWBSTRACTC_SCPLOWProtein-protein interactions are crucial to many biological processes, and predicting the effect of amino acid mutations on binding is important for protein engineering. While data-driven approaches using deep learning have shown promise, the scarcity of annotated experimental data remains a major challenge. In this work, we propose a new approach that predicts mutational effects on binding using the change in conformational flexibility of the protein-protein interface. Our approach, named Rotamer Density Estimator (RDE), employs a flow-based generative model to estimate the probability distribution of protein side-chain conformations and uses entropy to measure flexibility. RDE is trained solely on protein structures and does not require the supervision of experimental values of changes in binding affinities. Furthermore, the unsupervised representations extracted by RDE can be used for downstream neural network predictions with even greater accuracy. Our method outperforms empirical energy functions and other machine learning-based approaches.

bioinformatics↗