<div><div><div><p>We describe a Transformer model for a retrosynthetic reaction prediction task. The model is trained on 45 033 experimental reaction examples extracted from USA patents. It can successfully predict the reactants set for 42.7% of cases on the external test set. During the training procedure, we applied different learning rate schedules and snapshot learning. These techniques can prevent overfitting and thus can be a reason to get rid of internal validation dataset that is advantageous for deep models with millions of parameters. We thoroughly investigated different approaches to train Transformer models and found that snapshot learning with averaging weights on learning rates minima works best. While decoding the model output probabilities there is a strong influence of the temperature that improves at T=1.3 the accuracy of models up to 1-2%.</p></div></div></div>

SummaryThe spindle assembly checkpoint (SAC) is a refined surveillance mechanism which ensures that chromosomes undergoing mitosis do not segregate until they are properly attached to the spindle microtubules (MT). The SAC has been extensively studied in metazoans and yeast, but little is known about its role in plants.We identified proteins interacting with a MT-associated protein MAP65-3, which plays a critical role in organising mitotic MT arrays, and carried out a functional analysis of previously and newly identified SAC components.We show that Arabidopsis SAC proteins BUB3.1, MAD2, BUBR1/MAD3s and BRK1 interact with each other and with MAP65-3. We found that two BUBR1/MAD3s interacted specifically at centromeres. When stably expressed in Arabidopsis, BRK1 localised to the kinetochores during all stages of the mitotic cell cycle. Early in mitosis, BUB3.1 and BUBR1/MAD3.1 localise to the mitotic spindle, where MAP65-3 organises spindle MTs. A double-knockout mad3.1 mad3.2 mutant presented spindle MT abnormalities, chromosome misalignments on the metaphase plate and the production of lagging chromosomes and micronuclei during mitosis.We conclude that BRK1 and BUBR1/MAD3-related proteins play a key role in ensuring faithful chromosome segregation during mitosis and that their interaction with MAP65-3 may be important for the regulation of MT-chromosome attachment.

The study of the genome and the proteome of different species and representatives of distinct kingdoms, especially detection of proteome via wide-scaled analyses has various challenges and pitfalls. Attempts to combine all available information together and isolate some common features for determination of the pathway and their mechanism of action generally have a highly complicated nature. However, microtubule (MT) monomers are highly conserved protein structures, and microtubules are structurally conserved from Homo sapiens to Arabidopsis thaliana. The interaction of MT elements with microtubule-associated proteins and post-translational modifiers is fully dependent on protein interfaces, and almost all MT modifications are well described except acetylation. Crystallography and interactome data using different approaches were combined to identify conserved proteins important in acetylation of microtubules. Application of computational methods and comparative analysis of binding modes generated a robust predictive model of acetylation of the ϵ-amino group of Lys40 in α-tubulins. In turn, the model discarded some probable mechanisms of interaction between elements of interest. Reconstruction of unresolved protein structures was carried out with modeling by homology to the existing crystal structure (PDBID: 1Z2B) from B. taurus using Swiss-model server, followed by a molecular dynamics simulation. Docking of the human tubulin fragment with Lys40 into the active site of α-tubulin acetyltransferase, reproduces the binding mode of peptidomimetic from X-ray structure (PDBID: 4PK3).