Ribonanza: deep learning of RNA structure through dual crowdsourcing.

Prediction of RNA structure from sequence remains an unsolved problem, and progress has been slowed by a paucity of experimental data. Here, we present Ribonanza, a dataset of chemical mapping measurements on two million diverse RNA sequences collected through Eterna and other crowdsourced initiatives. Ribonanza measurements enabled solicitation, training, and prospective evaluation of diverse deep neural networks through a Kaggle challenge, followed by distillation into a single, self-contained model called RibonanzaNet. When fine tuned on auxiliary datasets, RibonanzaNet achieves state-of-the-art performance in modeling experimental sequence dropout, RNA hydrolytic degradation, and RNA secondary structure, with implications for modeling RNA tertiary structure.

High-Speed Atomic Force Microscopy Visualization of the Dynamics of the Multienzyme Fatty Acid Synthase.

Multienzymes, such as the protein metazoan fatty acid synthase (FAS), are giant and highly dynamic molecular machines for critical biosynthetic processes. The molecular architecture of FAS was elucidated by static high-resolution crystallographic analysis, while electron microscopy revealed large-scale conformational variability in FAS with some correlation to functional states in catalysis. However, little is known about time scales of conformational dynamics, the trajectory of motions in individual FAS molecules, and the extent of coupling between catalysis and structural changes. Here, we present an experimental single-molecule approach to film immobilized or selectively tethered FAS in solution at different viewing angles and high spatiotemporal resolution using high-speed atomic force microscopy. Mobility of individual regions of the multienzyme is recognized in video sequences, and correlation of shape features implies a convergence of temporal resolution and velocity of FAS dynamics. Conformational variety can be identified and grouped by reference-free 2D class averaging, enabling the tracking of conformational transitions in movies. The approach presented here is suited for comprehensive studies of the dynamics of FAS and other multienzymes in aqueous solution at the single-molecule level.

Mechanism of foreign DNA recognition by a CRISPR RNA-guided surveillance complex from Pseudomonas aeruginosa.

The Type I-F CRISPR-mediated (clustered regularly interspaced short palindromic repeats) adaptive immune system in Pseudomonas aeruginosa consists of two CRISPR loci and six CRISPR-associated (cas) genes. Foreign DNA surveillance is performed by a complex of Cas proteins (Csy1­4) that assemble with a CRISPR RNA (crRNA) into a 350-kDa ribonucleoprotein called the Csy complex. Here, we show that foreign nucleic acid recognition by the Csy complex proceeds through sequential steps, initiated by detection of two consecutive guanine­cytosine base pairs (G­C/G­C) located adjacent to the complementary DNA target. We show that this motif, called the PAM (protospacer adjacent motif), must be double-stranded and that single-stranded PAMs do not provide significant discriminating power. Binding assays performed with G­C/G­C-rich competitor sequences indicate that the Csy complex interacts directly with this dinucleotide motif, and kinetic analyses reveal that recognition of a G­C/G­C motif is a prerequisite for crRNA-guided binding to a target sequence. Together, these data indicate that the Csy complex first interacts with G­C/G­C base pairs and then samples adjacent target sequences for complementarity to the crRNA guide.

Evolutionary origins of the multienzyme architecture of giant fungal fatty acid synthase.

Fungal fatty acid synthase (fFAS) is a key paradigm for the evolution of complex multienzymes. Its 48 functional domains are embedded in a matrix of scaffolding elements, which comprises almost 50% of the total sequence and determines the emergent multienzymes properties of fFAS. Catalytic domains of fFAS are derived from monofunctional bacterial enzymes, but the evolutionary origin of the scaffolding elements remains enigmatic. Here, we identify two bacterial protein families of noncanonical fatty acid biosynthesis starter enzymes and trans-acting polyketide enoyl reductases (ERs) as potential ancestors of scaffolding regions in fFAS. The architectures of both protein families are revealed by representative crystal structures of the starter enzyme FabY and DfnA-ER. In both families, a striking structural conservation of insertions to scaffolding elements in fFAS is observed, despite marginal sequence identity. The combined phylogenetic and structural data provide insights into the evolutionary origins of the complex multienzyme architecture of fFAS.