TEMPRO: nanobody melting temperature estimation model using protein embeddings.

Single-domain antibodies (sdAbs) or nanobodies have received widespread attention due to their small size (~ 15 kDa) and diverse applications in bio-derived therapeutics. As many modern biotechnology breakthroughs are applied to antibody engineering and design, nanobody thermostability or melting temperature (Tm) is crucial for their successful utilization. In this study, we present TEMPRO which is a predictive modeling approach for estimating the Tm of nanobodies using computational methods. Our methodology integrates various nanobody biophysical features to include Evolutionary Scale Modeling (ESM) embeddings, NetSurfP3 structural predictions, pLDDT scores per sdAb region from AlphaFold2, and each sequence's physicochemical characteristics. This approach is validated with our combined dataset containing 567 unique sequences with corresponding experimental Tm values from a manually curated internal data and a recently published nanobody database, NbThermo. Our results indicate the efficacy of protein embeddings in reliably predicting the Tm of sdAbs with mean absolute error (MAE) of 4.03 °C and root mean squared error (RMSE) of 5.66 °C, thus offering a valuable tool for the optimization of nanobodies for various biomedical and therapeutic applications. Moreover, we have validated the models' performance using experimentally determined Tms from nanobodies not found in NbThermo. This predictive model not only enhances nanobody thermostability prediction, but also provides a useful perspective of using embeddings as a tool for facilitating a broader applicability of downstream protein analyses.

Local Control Model of a Human Ventricular Myocyte: An Exploration of Frequency-Dependent Changes and Calcium Sparks.

Calcium (Ca2+) sparks are the elementary events of excitation-contraction coupling, yet they are not explicitly represented in human ventricular myocyte models. A stochastic ventricular cardiomyocyte human model that adapts to intracellular Ca2+ ([Ca2+]i) dynamics, spark regulation, and frequency-dependent changes in the form of locally controlled Ca2+ release was developed. The 20,000 CRUs in this model are composed of 9 individual LCCs and 49 RyRs that function as couplons. The simulated action potential duration at 1 Hz steady-state pacing is ~0.280 s similar to human ventricular cell recordings. Rate-dependence experiments reveal that APD shortening mechanisms are largely contributed by the L-type calcium channel inactivation, RyR open fraction, and [Ca2+]myo concentrations. The dynamic slow-rapid-slow pacing protocol shows that RyR open probability during high pacing frequency (2.5 Hz) switches to an adapted "nonconducting" form of Ca2+-dependent transition state. The predicted force was also observed to be increased in high pacing, but the SR Ca2+ fractional release was lower due to the smaller difference between diastolic and systolic [Ca2+]SR. Restitution analysis through the S1S2 protocol and increased LCC Ca2+-dependent activation rate show that the duration of LCC opening helps modulate its effects on the APD restitution at different diastolic intervals. Ultimately, a longer duration of calcium sparks was observed in relation to the SR Ca2+ loading at high pacing rates. Overall, this study demonstrates the spontaneous Ca2+ release events and ion channel responses throughout various stimuli.

PepVAE: Variational Autoencoder Framework for Antimicrobial Peptide Generation and Activity Prediction.

New methods for antimicrobial design are critical for combating pathogenic bacteria in the post-antibiotic era. Fortunately, competition within complex communities has led to the natural evolution of antimicrobial peptide (AMP) sequences that have promising bactericidal properties. Unfortunately, the identification, characterization, and production of AMPs can prove complex and time consuming. Here, we report a peptide generation framework, PepVAE, based around variational autoencoder (VAE) and antimicrobial activity prediction models for designing novel AMPs using only sequences and experimental minimum inhibitory concentration (MIC) data as input. Sampling from distinct regions of the learned latent space allows for controllable generation of new AMP sequences with minimal input parameters. Extensive analysis of the PepVAE-generated sequences paired with antimicrobial activity prediction models supports this modular design framework as a promising system for development of novel AMPs, demonstrating controlled production of AMPs with experimental validation of predicted antimicrobial activity.