Deep Learning in Therapeutic Antibody Development.

Deep learning applied to antibody development is in its adolescence. Low data volumes and biological platform differences make it challenging to develop supervised models that can predict antibody behavior in actual commercial development steps. But successes in modeling general protein behaviors and early antibody models give indications of what is possible for antibodies in general, particularly since antibodies share a common fold. Meanwhile, new methods of data collection and the development of unsupervised and self-supervised deep learning methods like generative models and masked language models give the promise of rich and deep data sets and deep learning architectures for better supervised model development. Together, these move the industry toward improved developability , lower costs, and broader access of biotherapeutics .

Evaluation of Crystal Zenith Microtiter Plates for High-Throughput Formulation Screening.

Formulation screening for biotherapeutics can cover a vast array of excipients and stress conditions. These studies consume quantities of limited material and, with higher concentrated therapeutics, more material is needed. Here, we evaluate the use of crystal zenith (CZ) microtiter plates in conjunction with FluoroTec-coated butyl rubber mats as a small-volume, high-throughput system for formulation stability studies. The system was studied for evaporation, edge effects, and stability with comparisons to type 1 glass and CZ vials for multiple antibodies and formulations. Evaporation was minimal at 4°C and could be reduced at elevated temperatures using sealed, mylar bags. Edge effects were not observed until 12 weeks at 40°C. The overall stability ranking as measured by the rate of change in high molecular weight and percent main peak species was comparable across both vials and plates at 4°C and 40°C out to 12 weeks. Product quality attributes as measured by the multi-attribute method were also comparable across all containers for each molecule formulation. A potential difference was measured for subvisible particle analysis, with the plates measuring lower particle counts than the vials. Overall, CZ plates are a viable alternative to traditional vials for small-volume, high-throughput formulation stability screening studies.

Adapting the chemical unfolding assay for high-throughput protein screening using experimental and spectroscopic corrections.

The chemical unfolding (denaturation) assay can be used to calculate the change in the Gibbs free energy of unfolding, ΔG, and inflection point of unfolding, to collectively inform on molecule stability. Here, we evaluated methods for calculating the ΔG across 23 monoclonal antibody sequence variants. These methods are based on how the measured output (intrinsic fluorescence intensity) is treated, including utilizing (a) a single wavelength, (b) a ratio of two wavelengths, (c) a ratio of a single wavelength to an area, and (d) a scatter correction plus a ratio of a single wavelength to an area. When applied to the variants, the three ratio methods showed comparable results, with a similar pooled standard deviation for the ΔG calculation, while the single-wavelength method is shown as inadequate for the data in this study. However, when light scattering is introduced to simulated data, only the scatter-correction area normalization method proves robust. Using this method, common plate-based spectrophotometers found in many laboratories can be used for high-throughput screening of mAb variants and formulation stability studies.

Muscle oxygenation as an indicator of shock severity in patients with suspected severe sepsis or septic shock.

PURPOSE: The aim of this pilot study was to evaluate the potential of a new noninvasive optical measurement of muscle oxygenation (MOx) to identify shock severity in patients with suspected sepsis. METHODS: We enrolled 51 adult patients in the emergency department (ED) who presented with possible sepsis using traditional Systematic Inflammatory Response Syndrome criteria or who triggered a "Code Sepsis." Noninvasive MOx measurements were made from the first dorsal interosseous muscles of the hand once potential sepsis/septic shock was identified, as soon as possible after admission to the ED. Shock severity was defined by concurrent systolic blood pressure, heart rate, and serum lactate levels. MOx was also measured in a control group of 17 healthy adults. RESULTS: Mean (± SD) MOx in the healthy control group was 91.0 ± 5.5% (n = 17). Patients with mild, moderate, and severe shock had mean MOx values of 79.4 ± 21.2%, 48.6 ± 28.6%, and 42.2 ± 4.7%, respectively. Mean MOx for the mild and moderate shock severity categories were statistically different from healthy controls and from each other based on two-sample t-tests (p < 0.05). CONCLUSIONS: We demonstrate that noninvasive measurement of MOx was associated with clinical assessment of shock severity in suspected severe sepsis or septic shock. The ability of MOx to detect even mild septic shock has meaningful implications for emergency care, where decisions about triage and therapy must be made quickly and accurately. Future longitudinal studies may validate these findings and the value of MOx in monitoring patient status as treatment is administered.

Muscle oxygenation measurement in humans by noninvasive optical spectroscopy and Locally Weighted Regression.

We have developed a method to make real-time, continuous, noninvasive measurements of muscle oxygenation (Mox) from the surface of the skin. A key development was measurement in both the visible and near infrared (NIR) regions. Measurement of both oxygenated and deoxygenated myoglobin and hemoglobin resulted in a more accurate measurement of Mox than could be achieved with measurement of only the deoxygenated components, as in traditional near-infrared spectroscopy (NIRS). Using the second derivative with respect to wavelength reduced the effects of scattering on the spectra and also made oxygenated and deoxygenated forms more distinguishable from each other. Selecting spectral bands where oxygenated and deoxygenated forms absorb filtered out noise and spectral features unrelated to Mox. NIR and visible bands were scaled relative to each other in order to correct for errors introduced by normalization. Multivariate Curve Resolution (MCR) was used to estimate Mox from spectra within each data set collected from healthy subjects. A Locally Weighted Regression (LWR) model was built from calibration set spectra and associated Mox values from 20 subjects using 2562 spectra. LWR and Partial Least Squares (PLS) allow accurate measurement of Mox despite variations in skin pigment or fat layer thickness in different subjects. The method estimated Mox in five healthy subjects with an RMSE of 5.4%.

Estimation of the age of a weathered mixture of volatile organic compounds.

The ability to predict the amount of time that a light petroleum mixture has been weathered could have many applications, such as aiding forensic investigators in determining the cause and intent of a fire. In our study, an evaporation chamber that permits control of airflow and temperature was constructed and used to weather a model nine-component hydrocarbon mixture. The composition of the mixture was monitored over time by gas chromatography and a variety of chemometric models were explored, including partial least squares (PLS), nonlinear PLS (PolyPLS) and locally weighted regression (LWR or loess). A hierarchical application of multivariate techniques was able to predict the time for which a sample had been exposed to evaporative weathering. A classification model based on partial least squares discriminant analysis (PLS-DA) could predict whether a sample was relatively fresh (< 12 h exposure time) or highly weathered (>20 h exposure time). Subsequent regression models for these individual classes were evaluated for accuracy using the root mean square error of prediction (RMSEP). Prior to regression model calculation, y-gradient generalized least squares weighting (GLSW) was used to preprocess the data by removing variance from the X-block, which was orthogonal to the Y-block. LWR was found to be the most successful regression method, whereby fresh samples could be predicted to within 40 min of exposure and highly weathered samples predicted to within 5.6h. These results suggest that our hierarchical chemometric approach may also allow us to estimate the age of more complicated light petroleum mixtures, such as gasoline.

Multivariate curve resolution applied to infrared reflection measurements of soil contaminated with an organophosphorus analyte.

Multivariate curve resolution (MCR) is a powerful technique for extracting chemical information from measured spectra of complex mixtures. A modified MCR technique that utilized both measured and second-derivative spectra to account for observed sample-to-sample variability attributable to changes in soil reflectivity was used to estimate the spectrum of dibutyl phosphate (DBP) adsorbed on two different soil types. This algorithm was applied directly to measurements of reflection spectra of soils coated with analyte without resorting to soil preparations such as grinding or dilution in potassium bromide. The results provided interpretable spectra that can be used to guide strategies for detection and classification of organic analytes adsorbed on soil. Comparisons to the neat DBP liquid spectrum showed that the recovered analyte spectra from both soils showed spectral features from methyl, methylene, hydroxyl, and P=O functional groups, but most conspicuous was the absence of the strong PO-(CH2)3CH3 stretch absorption at 1033 cm(-1). These results are consistent with those obtained previously using extended multiplicative scatter correction.