Konczak, J. | bioRxiv Science

bioRxiv2024

Altered cortico-cerebellar connectivity in cerebellar degeneration patients improves with motor training

People with cerebellar degeneration show characteristic ataxic motor impairments. Despite cerebellar dysfunction, they can still improve motor performance through sensorimotor training. Yet, how such training affects functional brain networks affected by cerebellar degeneration is unknown. We here investigated neuroplastic changes in the cortico-cerebellar network after a five-day forearm movement training in 40 patients with mild to severe cerebellar degeneration and 40 age- and sex-matched healthy controls. Participants were assigned to one of four motor training conditions, varying online visual feedback and explicit verbal feedback. Anatomical and resting-state fMRI was collected on the days before and after training. To overcome the limitations of standard brain templates that fail in the presence of severe anatomical abnormalities, we developed a specific template for comparing cerebellar patients with age-matched controls. Our new template reduced the spatial spread of cerebellar anatomical landmarks by 30% relative to existing templates and tripled fMRI noise classification accuracy. Using this pipeline, we found that patients showed impaired connectivity between cerebellar motor regions and neocortical visuomotor and premotor regions at baseline compared to controls, whereas their cortico-cortical connectivity remained intact. Training with vision strengthened connectivity in the cortico-cerebellar visuomotor network contralateral to the trained arm in all participants. Cerebellar patients exhibited additional increased connectivity ipsilateral to the training arm in this network. Further, training with explicit verbal feedback facilitated connectivity between a cerebellar cognitive region and dorsolateral prefrontal cortex. These results indicate that motor training in cerebellar degeneration leads to enhanced functional connectivity of the cortico-cerebellar network.

neuroscience↗

bioRxiv2022

Structural pre-training improves physical accuracy of antibody structure prediction using deep learning.

Protein folding problem obtained a practical solution recently, owing to advances in deep learning. There are classes of proteins though, such as antibodies, that are structurally unique, where the general solution still lacks. In particular, the prediction of the CDR-H3 loop, which is an instrumental part of an antibody in its antigen recognition abilities, remains a challenge. Antibody-specific deep learning frameworks were proposed to tackle this problem noting great progress, both on accuracy and speed fronts. Oftentimes though, the original networks produce physically implausible bond geometries that then need to undergo a time-consuming energy minimization process. Here we hypothesized that pre-training the network on a large, augmented set of models with correct physical geometries, rather than a small set of real antibody X-ray structures, would allow the network to learn better bond geometries. We show that fine-tuning such a pre-trained network on a task of shape prediction on real X-ray structures improves the number of correct peptide bond distances. We further demonstrate that pre-training allows the network to produce physically plausible shapes on an artificial set of CDR-H3s, showing the ability to generalize to the vast antibody sequence space. We hope that our strategy will benefit the development of deep learning antibody models that rapidly generate physically plausible geometries, without the burden of time-consuming energy minimization.

bioinformatics↗