Binding peptide generation for MHC Class I proteins with deep reinforcement learning.

MOTIVATION: MHC Class I protein plays an important role in immunotherapy by presenting immunogenic peptides to anti-tumor immune cells. The repertoires of peptides for various MHC Class I proteins are distinct, which can be reflected by their diverse binding motifs. To characterize binding motifs for MHC Class I proteins, in vitro experiments have been conducted to screen peptides with high binding affinities to hundreds of given MHC Class I proteins. However, considering tens of thousands of known MHC Class I proteins, conducting in vitro experiments for extensive MHC proteins is infeasible, and thus a more efficient and scalable way to characterize binding motifs is needed. RESULTS: We presented a de novo generation framework, coined PepPPO, to characterize binding motif for any given MHC Class I proteins via generating repertoires of peptides presented by them. PepPPO leverages a reinforcement learning agent with a mutation policy to mutate random input peptides into positive presented ones. Using PepPPO, we characterized binding motifs for around 10 000 known human MHC Class I proteins with and without experimental data. These computed motifs demonstrated high similarities with those derived from experimental data. In addition, we found that the motifs could be used for the rapid screening of neoantigens at a much lower time cost than previous deep-learning methods. AVAILABILITY AND IMPLEMENTATION: The software can be found in https://github.com/minrq/pMHC. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Data Sanitization to Reduce Private Information Leakage from Functional Genomics.

The generation of functional genomics datasets is surging, because they provide insight into gene regulation and organismal phenotypes (e.g., genes upregulated in cancer). The intent behind functional genomics experiments is not necessarily to study genetic variants, yet they pose privacy concerns due to their use of next-generation sequencing. Moreover, there is a great incentive to broadly share raw reads for better statistical power and general research reproducibility. Thus, we need new modes of sharing beyond traditional controlled-access models. Here, we develop a data-sanitization procedure allowing raw functional genomics reads to be shared while minimizing privacy leakage, enabling principled privacy-utility trade-offs. Our protocol works with traditional Illumina-based assays and newer technologies such as 10x single-cell RNA sequencing. It involves quantifying the privacy leakage in reads by statistically linking study participants to known individuals. We carried out these linkages using data from highly accurate reference genomes and more realistic environmental samples.

An integrative ENCODE resource for cancer genomics.

ENCODE comprises thousands of functional genomics datasets, and the encyclopedia covers hundreds of cell types, providing a universal annotation for genome interpretation. However, for particular applications, it may be advantageous to use a customized annotation. Here, we develop such a custom annotation by leveraging advanced assays, such as eCLIP, Hi-C, and whole-genome STARR-seq on a number of data-rich ENCODE cell types. A key aspect of this annotation is comprehensive and experimentally derived networks of both transcription factors and RNA-binding proteins (TFs and RBPs). Cancer, a disease of system-wide dysregulation, is an ideal application for such a network-based annotation. Specifically, for cancer-associated cell types, we put regulators into hierarchies and measure their network change (rewiring) during oncogenesis. We also extensively survey TF-RBP crosstalk, highlighting how SUB1, a previously uncharacterized RBP, drives aberrant tumor expression and amplifies the effect of MYC, a well-known oncogenic TF. Furthermore, we show how our annotation allows us to place oncogenic transformations in the context of a broad cell space; here, many normal-to-tumor transitions move towards a stem-like state, while oncogene knockdowns show an opposing trend. Finally, we organize the resource into a coherent workflow to prioritize key elements and variants, in addition to regulators. We showcase the application of this prioritization to somatic burdening, cancer differential expression and GWAS. Targeted validations of the prioritized regulators, elements and variants using siRNA knockdowns, CRISPR-based editing, and luciferase assays demonstrate the value of the ENCODE resource.

Combining Molecular and Spin Dynamics Simulations with Solid-State NMR: A Case Study of Amphiphilic Lysine-Leucine Repeat Peptide Aggregates.

Interpreting dynamics in solid-state molecular systems requires characterization of the potentially heterogeneous environmental contexts of molecules. In particular, the analysis of solid-state nuclear magnetic resonance (ssNMR) data to elucidate molecular dynamics (MD) involves modeling the restriction to overall tumbling by neighbors, as well as the concentrations of water and buffer. In this exploration of the factors that influence motion, we utilize atomistic MD trajectories of peptide aggregates with varying hydration to mimic an amorphous solid-state environment and predict ssNMR relaxation rates. We also account for spin diffusion in multiply spin-labeled (up to 19 nuclei) residues, with several models of dipolar-coupling networks. The framework serves as a general approach to determine essential spin couplings affecting relaxation, benchmark MD force fields, and reveal the hydration dependence of dynamics in a crowded environment. We demonstrate the methodology on a previously characterized amphiphilic 14-residue lysine-leucine repeat peptide, LKα14 (Ac-LKKLLKLLKKLLKL-c), which has an α-helical secondary structure and putatively forms leucine-burying tetramers in the solid state. We measure the R1 relaxation rates of uniformly 13C-labeled and site-specific 2H-labeled leucines in the hydrophobic core of LKα14 at multiple hydration levels. Studies of 9 and 18 tetramer bundles reveal the following: (a) for the incoherent component of 13C relaxation, the nearest-neighbor spin interactions dominate, while the 1H-1H interactions have minimal impact; (b) the AMBER ff14SB dihedral barriers for the leucine Cγ-Cδ bond ("methyl rotation barriers") must be lowered by a factor of 0.7 to better match the 2H data; (c) proton-driven spin diffusion explains some of the discrepancy between experimental and simulated rates for the Cß and Cα nuclei; and (d) 13C relaxation rates are mostly underestimated in the MD simulations at all hydrations, and the discrepancies identify likely motions missing in the 50 ns MD trajectories.

A REDOR ssNMR Investigation of the Role of an N-Terminus Lysine in R5 Silica Recognition.

Diatoms are unicellular algae that construct cell walls called frustules by the precipitation of silica, using special proteins that order the silica into a wide variety of nanostructures. The diatom species Cylindrotheca fusiformis contains proteins called silaffins within its frustules, which are believed to assemble into supramolecular matrices that serve as both accelerators and templates for silica deposition. Studying the properties of these biosilicification proteins has allowed the design of new protein and peptide systems that generate customizable silica nanostructures, with potential generalization to other mineral systems. It is essential to understand the mechanisms of aggregation of the protein and its coprecipitation with silica. We continue previous investigations into the peptide R5, derived from silaffin protein sil1p, shown to independently catalyze the precipitation of silica nanospheres in vitro. We used the solid-state NMR technique 13C{29Si} and 15N{29Si} REDOR to investigate the structure and interactions of R5 in complex with coprecipitated silica. These experiments are sensitive to the strength of magnetic dipole-dipole interactions between the 13C nuclei in R5 and the 29Si nuclei in the silica and thus yield distance between parts of R5 and 29Si in silica. Our data show strong interactions and short internuclear distances of 3.74 ± 0.20 Å between 13C═O Lys3 and silica. On the other hand, the Cα and Cß nuclei show little or no interaction with 29Si. This selective proximity between the K3 C═O and the silica supports a previously proposed mechanism of rapid silicification of the antimicrobial peptide KSL (KKVVFKVKFK) through an imidate intermediate. This study reports for the first time a direct interaction between the N-terminus of R5 and silica, leading us to believe that the N-terminus of R5 is a key component in the molecular recognition process and a major factor in silica morphogenesis.

A study of phenylalanine side-chain dynamics in surface-adsorbed peptides using solid-state deuterium NMR and rotamer library statistics.

Extracellular matrix proteins adsorbed onto mineral surfaces exist in a unique environment where the structure and dynamics of the protein can be altered profoundly. To further elucidate how the mineral surface impacts molecular properties, we perform a comparative study of the dynamics of nonpolar side chains within the mineral-recognition domain of the biomineralization protein salivary statherin adsorbed onto its native hydroxyapatite (HAP) mineral surface versus the dynamics displayed by the native protein in the hydrated solid state. Specifically, the dynamics of phenylalanine side chains (viz., F7 and F14) located in the surface-adsorbed 15-amino acid HAP-recognition fragment (SN15: DpSpSEEKFLRRIGRFG) are studied using deuterium magic angle spinning ((2)H MAS) line shape and spin-lattice relaxation measurements. (2)H NMR MAS spectra and T1 relaxation times obtained from the deuterated phenylalanine side chains in free and HAP-adsorbed SN15 are fitted to models where the side chains are assumed to exchange between rotameric states and where the exchange rates and a priori rotameric state populations are varied iteratively. In condensed proteins, phenylalanine side-chain dynamics are dominated by 180° flips of the phenyl ring, i.e., the "π flip". However, for both F7 and F14, the number of exchanging side-chain rotameric states increases in the HAP-bound complex relative to the unbound solid sample, indicating that increased dynamic freedom accompanies introduction of the protein into the biofilm state. The observed rotameric exchange dynamics in the HAP-bound complex are on the order of 5-6 × 10(6) s(-1), as determined from the deuterium MAS line shapes. The dynamics in the HAP-bound complex are also shown to have some solution-like behavioral characteristics, with some interesting deviations from rotameric library statistics.

Silica morphogenesis by lysine-leucine peptides with hydrophobic periodicity.

The use of biomimetic approaches in the production of inorganic nanostructures is of great interest to the scientific and industrial community due to the relatively moderate physical conditions needed. In this vein, taking cues from silaffin proteins used by unicellular diatoms, several studies have identified peptide candidates for the production of silica nanostructures. In the current article, we study intensively one such silica-precipitating peptide, LKα14 (Ac-LKKLLKLLKKLLKL-c), an amphiphilic lysine/leucine repeat peptide that self-organizes into an α-helical secondary structure under appropriate concentration and buffer conditions. The suggested mechanism of precipitation is that the sequestration of hydrophilic lysines on one side of this helix allows interaction with the negatively charged surface of silica nanoparticles, which in turn can aggregate further into larger structures. To investigate the process, we carry out 1D and 2D solid-state NMR (ssNMR) studies on samples with one or two uniformly (13)C- and (15)N-labeled residues to determine the backbone and side-chain chemical shifts. We also further study the dynamics of two leucine residues in the sequence through (13)C spin-lattice relaxation times (T1) to determine the impact of silica coprecipitation on their mobility. Our results confirm the α-helical secondary structure in both the neat and silica-complexed states of the peptide, and the patterns of chemical shift and relaxation time changes between the two states suggest possible mechanisms of self-aggregation and silica precipitation.