Evaluation of rapid evaporative ionization mass spectrometry (REIMS) for the prediction of slice shear force and biochemical markers of tenderness in beef Longissimus lumborum steaks
Bibliographic record
Abstract
The objective of this study was to evaluate rapid evaporative ionization mass spectrometry (REIMS) as a rapid method to predict slice shear force (SSF) and biochemical markers of beef tenderness. Steak samples were randomly collected from beef carcasses (Canada AA, n = 1505; Canada AAA, n = 1363) over a three-year period. Steaks were aged for 14 d, then tenderness was determined using SSF. Metabolomic profiling of beef samples was performed using REIMS (N = 2,853). A subset of samples (n = 600) were selected to evaluate sarcomere length, myofibril fragmentation index (MFI), desmin, and troponin-T degradation. Thirteen machine learning algorithms were used to build several predictive models. Data were reduced using feature selection (FS) and principal component analysis followed by FS (PCA-FS). No models could predict SSF tenderness category with a higher accuracy than the no information rate (NIR, 59.5%) for FS and PCA-FS datasets (P ≥ 0.05). Population mean and standard deviation (SD) were used to generate 4 SD categories (± 2) for further predictions. No models could predict SD category with a higher accuracy than the NIR using the FS dataset (P > 0.05). Accuracies to predict SD category using the PCA-FS dataset ranged from 55.0% to 83.0%. Top accuracies of 82.8% and 83.0% were generated from the treebag and random forest (RF) algorithms (NIR = 55.0%, P < 0.001). Accuracies to predict quality grade using the FS dataset ranged from 52.5% to 85.3%. Top accuracies of 84.6% and 85.3% were generated from SVM Radial and XGBoost, respectively (NIR = 52.5%, P < 0.001). Using the PCA-FS dataset, all models could predict quality grade with a higher accuracy than the NIR (P < 0.001). The top accuracies of 82.8% and 84.2% were generated from SVM Radial and RF (P < 0.001). A stepwise regression model was built to evaluate the relationship between SSF values and the spectra data generated from REIMS (N = 2,853). The selected REIMS bins accounted for 7.2% of the variation in predicted SSF value (R2 = 0.072; P < 0.001). Stepwise regression models were built to evaluate the relationship between sarcomere length, MFI, intact desmin, degraded desmin, intact troponin- T, and degraded troponin- T and the spectra data generated from REIMS (n = 600). The selected REIMS bins accounted for 64.4% of the variation in predicted sarcomere length (R2 = 0.644, P < 0.001), 31.6% in predicted MFI (R2 = 0.316, P < 0.001), 58.3% in predicted intact desmin (R2 = 0.583, P < 0.001), 50.9% in predicted degraded desmin (42kDa) (R2 = 0.509, P < 0.001), 54.2% in predicted degraded desmin (38kDa) (R2 = 0.542, P < 0.001), 12.8% in predicted intact troponin- T (R2 = 0.128, P < 0.001), 6.9% in predicted degraded troponin- T (30kDa) (R2 = 0.069, P < 0.001), and 8.4% in predicted degraded troponin- T (28kDa) (R2 = 0.084, P < 0.001). Segregation of samples based on SSF, sarcomere length, MFI, desmin, and troponin- T degradation was performed using K- means clustering, resulting in 3 clusters. The top accuracy using the FS dataset to predict cluster was 50.9%, generated from the XGBoost algorithm (NIR = 41.5%, P = 0.03). No models could predict k- means cluster with a higher accuracy than the NIR using the PCA-FS dataset (P > 0.05 for all models). Overall, REIMS showed an ability to predict the most tender and toughest steaks with a relatively high degree of accuracy. The RF and Treebag algorithms performed well in identifying tenderness and quality grade, so these algorithms could be further developed to improve prediction accuracies, allowing REIMS to be used as a rapid assessment of carcass quality.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".