Predicting the poses of small-molecule ligands in protein binding sites is often done by virtual screening algorithms such as DOCK. In principle, molecular dynamics (MD) using atomistic force fields could give better free-energy-based pose selection, but MD is computationally expensive. Here, we ask if modeling employing limited data (MELD)-accelerated MD (MELD × MD) can pick out the best DOCK poses taken as input. We study 30 different ligand-protein pairs. MELD × MD finds native poses, based on best free energies, in 23 out of the 30 cases, 20 of which were previously known DOCK failures. We conclude that MELD × MD can add value for predicting accurate poses of small molecules bound to proteins.

Cellular reprogramming is a promising technology to develop disease models and cell-based therapies. Identification of the key regulators defining the cell type specificity is pivotal to devising reprogramming cocktails for successful cell conversion but remains a great challenge. Here, we present a systems biology approach called Taiji-reprogram to efficiently uncover transcription factor (TF) combinations for conversion between 154 diverse cell types or tissues. This method integrates the transcriptomic and epigenomic data to construct cell-type specific genetic networks and assess the global importance of TFs in the network. Comparative analysis across cell types revealed TFs that are specifically important in a particular cell type and often tightly associated with cell-type specific functions. A systematic search of TFs with differential importance in the source and target cell types uncovered TF combinations for desired cell conversion. We have shown that Taiji-reprogram outperformed the existing methods to better recover the TFs in the experimentally validated reprogramming cocktails. This work not only provides a comprehensive catalog of TFs defining cell specialization but also suggests TF combinations for direct cell conversion.

Protein-protein interactions (PPIs) are important for many biological processes, but predicting them from sequence data remains challenging. Existing deep learning models often cannot generalize to proteins not present in the training set and do not provide uncertainty estimates for their predictions. To address these limitations, we present TUnA, a Transformer-based uncertainty-aware model for PPI prediction. TUnA uses ESM-2 embeddings with Transformer encoders and incorporates a Spectral-normalized Neural Gaussian Process. TUnA achieves state-of-the-art performance and, importantly, evaluates uncertainty for unseen sequences. We demonstrate that TUnAs uncertainty estimates can effectively identify the most reliable predictions, significantly reducing false positives. This capability is crucial in bridging the gap between computational predictions and experimental validation.

How the locus-specificity of epigenetic modifications is regulated remains an unanswered question. A contributing mechanism is that epigenetic enzymes are recruited to specific loci by DNA binding factors recognizing particular sequence motifs (referred to as epi-motifs). Using these motifs to predict biological outputs depending on local epigenetic state such as somatic mutation rates would confirm their functionality. Here, we used DNA motifs including known TF motifs and epi-motifs as a surrogate of epigenetic signals to predict somatic mutation rates in 13 cancers at an average 23kbp resolution. We implemented an interpretable neural network model, called contextual regression, to successfully learn the universal relationship between mutations and DNA motifs, and uncovered motifs that are most impactful on the regional mutation rates such as TP53 and epi-motifs associated with H3K9me3. Furthermore, we identified genomic regions with significantly higher mutation rates than the expected values in each individual tumor and demonstrated that such cancer-related regions can accurately predict cancer types. Interestingly, we found that the same mutation signatures often have different contributions to cancer-related and cancer-independent regions, and we also identified the motifs with the most contribution to each mutation signature.

The amyloid protein aggregation associated with diseases such as Alzheimer's, Parkinson's and type II diabetes (among many others) features a bewildering variety of β-sheet-rich structures in transition from native proteins to ordered oligomers and fibres. The variation in the amino-acid sequences of the β-structures presents a challenge to developing a model system of β-sheets for the study of various amyloid aggregates. Here, we introduce a family of robust β-sheet macrocycles that can serve as a platform to display a variety of heptapeptide sequences from different amyloid proteins. We have tailored these amyloid β-sheet mimics (ABSMs) to antagonize the aggregation of various amyloid proteins, thereby reducing the toxicity of amyloid aggregates. We describe the structures and inhibitory properties of ABSMs containing amyloidogenic peptides from the amyloid-β peptide associated with Alzheimer's disease, β(2)-microglobulin associated with dialysis-related amyloidosis, α-synuclein associated with Parkinson's disease, islet amyloid polypeptide associated with type II diabetes, human and yeast prion proteins, and Tau, which forms neurofibrillary tangles.

New materials capable of binding carbon dioxide are essential for addressing climate change. Here, we demonstrate that amyloids, self-assembling protein fibers, are effective for selective carbon dioxide capture. Solid-state NMR proves that amyloid fibers containing alkylamine groups reversibly bind carbon dioxide via carbamate formation. Thermodynamic and kinetic capture-and-release tests show the carbamate formation rate is fast enough to capture carbon dioxide by dynamic separation, undiminished by the presence of water, in both a natural amyloid and designed amyloids having increased carbon dioxide capacity. Heating to 100 °C regenerates the material. These results demonstrate the potential of amyloid fibers for environmental carbon dioxide capture.

We describe the performance of MELD-accelerated molecular dynamics (MELDxMD) in determining protein structures in the NMR-data-assisted category in CASP13. Seeded from web server predictions, MELDxMD was found best in the NMR category, over 17 targets, outperforming the next-best groups by a factor of ~4 in z-score. MELDxMD gives ensembles, not single structures; succeeds on a 326-mer, near the current upper limit for NMR structures; and predicts structures that match experimental residual dipolar couplings even though the only NMR-derived data used in the simulations was NOE-based ambiguous atom-atom contacts and backbone dihedrals. MELD can use noisy and ambiguous experimental information to reduce the MD search space. We believe MELDxMD is a promising method for determining protein structures from NMR data.

Amyloid protein aggregates are associated with dozens of devastating diseases including Alzheimer's, Parkinson's, ALS, and diabetes type 2. While structure-based discovery of compounds has been effective in combating numerous infectious and metabolic diseases, ignorance of amyloid structure has hindered similar approaches to amyloid disease. Here we show that knowledge of the atomic structure of one of the adhesive, steric-zipper segments of the amyloid-beta (Aβ) protein of Alzheimer's disease, when coupled with computational methods, identifies eight diverse but mainly flat compounds and three compound derivatives that reduce Aβ cytotoxicity against mammalian cells by up to 90%. Although these compounds bind to Aβ fibers, they do not reduce fiber formation of Aβ. Structure-activity relationship studies of the fiber-binding compounds and their derivatives suggest that compound binding increases fiber stability and decreases fiber toxicity, perhaps by shifting the equilibrium of Aβ from oligomers to fibers. DOI:http://dx.doi.org/10.7554/eLife.00857.001.

While modelling studies suggest that mesoscale eddies strengthen the subduction of mode waters, this eddy effect has never been observed in the field. Here we report results from a field campaign from March 2014 that captured the eddy effects on mode-water subduction south of the Kuroshio Extension east of Japan. The experiment deployed 17 Argo floats in an anticyclonic eddy (AC) with enhanced daily sampling. Analysis of over 3,000 hydrographic profiles following the AC reveals that potential vorticity and apparent oxygen utilization distributions are asymmetric outside the AC core, with enhanced subduction near the southeastern rim of the AC. There, the southward eddy flow advects newly ventilated mode water from the north into the main thermocline. Our results show that subduction by eddy lateral advection is comparable in magnitude to that by the mean flow--an effect that needs to be better represented in climate models.

Recently, predicting the native structures of proteins has become possible using computational molecular physics (CMP)─physics-based force fields sampled with proper statistics─but only for small proteins. Algorithms with better scaling are needed. We describe ML x MELD x MD, a molecular dynamics (MD) method that inputs residue contacts derived from machine learning (ML) servers into MELD, a Bayesian accelerator that preserves detailed-balance statistics. Contacts are derived from trRosetta-predicted distance histograms (distograms) and are integrated into MELD's atomistic MD as spatial restraints through parametrized potential functions. In the CASP14 blind prediction event, ML x MELD x MD predicted 13 native structures to better than 4.5 Å error, including for 10 proteins in the range of 115-250 amino acids long. Also, the scaling of simulation time vs protein length is much better than unguided MD: t_sim ∼ e^0.023N for ML x MELD x MD vs t_sim ∼ e^0.168N for MD alone. This shows how machine learning information can be leveraged to advance physics-based modeling of proteins.