Applications and challenges in designing VHH-based bispecific antibodies: leveraging machine learning solutions.

The development of bispecific antibodies that bind at least two different targets relies on bringing together multiple binding domains with different binding properties and biophysical characteristics to produce a drug-like therapeutic. These building blocks play an important role in the overall quality of the molecule and can influence many important aspects from potency and specificity to stability and half-life. Single-domain antibodies, particularly camelid-derived variable heavy domain of heavy chain (VHH) antibodies, are becoming an increasingly popular choice for bispecific construction due to their single-domain modularity, favorable biophysical properties, and potential to work in multiple antibody formats. Here, we review the use of VHH domains as building blocks in the construction of multispecific antibodies and the challenges in creating optimized molecules. In addition to exploring traditional approaches to VHH development, we review the integration of machine learning techniques at various stages of the process. Specifically, the utilization of machine learning for structural prediction, lead identification, lead optimization, and humanization of VHH antibodies.

Comparative effects of three different ventilatory treatments on arterial blood gas values and oxygen extraction in healthy anaesthetized dogs.

OBJECTIVE: To compare the effect of invasive continuous positive airway pressure (CPAP), pressure-controlled ventilation (PCV) with positive end-expiratory pressure (PEEP) and spontaneous breathing (SB) on PaO2, PaCO2 and arterial to central venous oxygen content difference (CaO2-CcvO2) in healthy anaesthetized dogs. STUDY DESIGN: Prospective randomized crossover study. ANIMALS: A group of 15 adult male dogs undergoing elective orchidectomy. METHODS: Dogs were anaesthetized [buprenorphine, medetomidine, propofol and isoflurane in an air oxygen (FiO2= 0.5)]. All ventilatory treatments (CPAP: 4 cmH2O; PCV: 10 cmH2O driving pressure; PEEP, 4 cmH2O; respiratory rate of 10 breaths minute-1 and inspiratory-to-expiratory ratio of 1:2; SB: no pressure applied) were applied in a randomized order during the same anaesthetic. Arterial and central venous blood samples were collected immediately before the start and at 20 minutes after each treatment. Data were compared using a general linear mixed model (p < 0.05). RESULTS: Median PaO2 was significantly higher after PCV [222 mmHg (29.6 kPa)] than after CPAP [202 mmHg (26.9 kPa)] and SB [208 mmHg (27.7 kPa)] (p < 0.001). Median PaCO2 was lower after PCV [48 mmHg (6.4 kPa)] than after CPAP [58 mmHg (7.7 kPa)] and SB [56 mmHg (7.5 kPa)] (p < 0.001). Median CaO2-CcvO2 was greater after PCV (4.36 mL dL-1) than after CPAP (3.41 mL dL-1) and SB (3.23 mL dL-1) (p < 0.001). PaO2, PaCO2 and CaO2-CcvO2 were no different between CPAP and SB (p > 0.99, p = 0.697 and p = 0.922, respectively). CONCLUSIONS AND CLINICAL RELEVANCE: CPAP resulted in similar arterial oxygenation, CO2 elimination and tissue oxygen extraction to SB. PCV resulted in improved arterial oxygenation and CO2 elimination. Greater oxygen extraction occurred with PCV than with CPAP and SB, offsetting its advantage of improved arterial oxygenation. The benefit of invasive CPAP over SB in the healthy anaesthetized dog remains uncertain.

Immobilization of African buffaloes (Syncerus caffer) using etorphine-midazolam compared with etorphine-azaperone.

OBJECTIVE: To compare induction times and physiological effects of etorphine-azaperone with etorphine-midazolam immobilization in African buffaloes. STUDY DESIGN: Randomized crossover study. ANIMALS: A group of 10 adult buffalo bulls (mean body weight 353 kg). METHODS: Etorphine-azaperone (treatment EA; 0.015 and 0.15 mg kg-1, respectively) and etorphine-midazolam (treatment EM; 0.015 and 0.15 mg kg-1, respectively) were administered once to buffaloes, 1 week apart. Once in sternal recumbency, buffaloes were instrumented and physiological variables recorded at 5 minute intervals, from 5 minutes to 20 minutes. Naltrexone (20 mg mg-1 etorphine dose) was administered intravenously at 40 minutes. Induction (dart placement to recumbency) and recovery (naltrexone administration to standing) times were recorded. Arterial blood samples were analysed at 5 and 20 minutes. Physiological data were compared between treatments using a general linear mixed model and reported as mean ± standard deviation. Time data were compared using Mann-Whitney U test and reported as median (interquartile range) with p ≤ 0.05. RESULTS: Actual drug doses administered for etorphine, azaperone and midazolam were 0.015 ± 0.001, 0.15 ± 0.01 and 0.16 ± 0.02 mg kg-1, respectively. Induction time for treatment EA was 3.3 (3.6) minutes and not different from 3.2 (3.2) minutes for treatment EM. The overall mean arterial blood pressure was significantly lower for treatment EA (102 ± 25 mmHg) than that for treatment EM (163 ± 18 mmHg) (p < 0.001). The PaO2 for treatment EA (37 ± 12 mmHg; 5.0 ± 1.6 kPa) was not different from that for treatment EM (43 ± 8 mmHg; 5.8 ± 1.1 kPa). Recovery time was 0.8 (0.6) minutes for treatment EA and did not differ from 1.1 (0.6) minutes for treatment EM. CONCLUSIONS AND CLINICAL RELEVANCE: Treatment EA was as effective as treatment EM for immobilization in this study. However, systemic arterial hypertension was a concern with treatment EM, and both combinations produced clinically relevant hypoxaemia. Supplemental oxygen administration is recommended with both drug combinations.

Patient-specific prediction of functional recovery after stroke.

Background and aims Clinical predictive models for stroke recovery could offer the opportunity of targeted early intervention and more specific information for patients and carers. In this study, we developed and validated a patient-specific prognostic model for monitoring recovery after stroke and assessed its clinical utility. Methods Four hundred and ninety-five patients from the population-based South London Stroke Register were included in a substudy between 2002 and 2004. Activities of daily living were assessed using Barthel Index) at one, two, three, four, six, eight, 12, 26, and 52 weeks after stroke. Penalized linear mixed models were developed to predict patients' functional recovery trajectories. An external validation cohort included 1049 newly registered stroke patients between 2005 and 2011. Prediction errors on discrimination and calibration were assessed. The potential clinical utility was evaluated using prognostic accuracy measurements and decision curve analysis. Results Predictive recovery curves showed good accuracy, with root mean squared deviation of 3 Barthel Index points and a R2 of 83% up to one year after stroke in the external cohort. The negative predictive values of the risk of poor recovery (Barthel Index <8) at three and 12 months were also excellent, 96% (95% CI [93.6-97.4]) and 93% [90.8-95.3], respectively, with a potential clinical utility measured by likelihood ratios (LR+:17 [10.8-26.8] at three months and LR+:11 [6.5-17.2] at 12 months). Decision curve analysis showed an increased clinical benefit, particularly at threshold probabilities of above 5% for predictive risk of poor outcomes. Conclusions A recovery curves tool seems to accurately predict progression of functional recovery in poststroke patients.