Predicting subacute ruminal acidosis from milk mid-infrared estimated fatty acids and machine learning on Canadian commercial dairy herds.

Our objective was to validate the possibility of detecting SARA from milk Fourier transform mid-infrared spectroscopy estimated fatty acids (FA) and machine learning. Subacute ruminal acidosis is a common condition in modern commercial dairy herds for which the diagnostic remains challenging due to its symptoms often being subtle, nonexclusive, and not immediately apparent. This observational study aimed at evaluating the possibility of predicting SARA by developing machine learning models to be applied to farm data and to provide an estimated portrait of SARA prevalence in commercial dairy herds. A first data set composed of 488 milk samples of 67 cows (initial DIM = 8.5 ± 6.18; mean ± SD) from 7 commercial dairy farms and their corresponding SARA classification (SARA+ if rumen pH <6.0 for 300 min, else SARA-) was used for the development of machine learning models. Three sets of predictive variables: i) milk major components (MMC), ii) milk FA (MFA), and iii) MMC combined with MFA (MMCFA) were submitted to 3 different algorithms, namely Elastic net (EN), Extreme gradient boosting (XGB), and Partial least squares (PLS), and evaluated using 3 different scenarios of cross-validation. Accuracy, sensitivity, and specificity of the resulting 27 models were analyzed using a linear mixed model. Model performance was not significantly affected by the choice of algorithm. Model performance was improved by including fatty acids estimations (MFA and MMCFA as opposed to MMC alone). Based on these results, one model was selected (algorithm: EN; predictive variables: MMCFA; 60.4, 65.4, and 55.3% of accuracy, sensitivity, and specificity, respectively) and applied to a large data set comprising the first test-day record (milk major components and FA within the first 70 DIM of 211,972 Holstein cows (219,503 samples) collected from 3001 commercial dairy herds. Based on this analysis, the within-herd SARA prevalence of commercial farms was estimated at 6.6 ± 5.29% ranging from 0 to 38.3%. A subsequent linear mixed model was built to investigate the herd-level factors associated to higher within-herd SARA prevalence. Milking system, proportion of primiparous cows, herd size and seasons were all herd-level factors affecting SARA prevalence. Furthermore, milk production was positively, and milk fat yield negatively associated with SARA prevalence. Due to their moderate levels of accuracy, the SARA prediction models developed in our study, using data from continuous pH measurements on commercial farms, are not suitable for diagnostic purpose. However, these models can provide valuable information at the herd level.

Relationship between farm management strategies, reticuloruminal pH variations, and risks of subacute ruminal acidosis.

Low reticuloruminal pH (rpH), often observed in subacute ruminal acidosis (SARA), may negatively affect rumen health and animal performance. To investigate the variability of rpH and the prevalence of SARA on commercial farms, we conducted an observational study on 110 early-lactation Holstein cows of different parities from 12 farms selected to cover a broad range of farm management strategies. The rpH of each cow was continuously monitored for 50 d using wireless boluses. To study the effects of animal and farm management characteristics on rpH, we used a multivariable mixed model analysis with the animal and farm as random effects. Automatic milking system and presence of corn silage in the ration were associated with a decrease in rpH of 0.37 and 0.20 pH units, respectively, whereas monensin supplementation was associated with an increase of 0.27 pH units. The rpH increased by 0.15 pH units during the first 60 d in milk. We defined a SARA-positive day as rpH below 5.8 (SARA5.8) or 6.0 (SARA6.0) for at least 300 min for 1 d. Using those definitions, during our study, a total of 38 (35%) and 65 (59%) cows experienced at least one episode of SARA5.8 and SARA6.0, respectively. The proportion of cows with at least one SARA-positive day varied among farms from 0 to 100%. Automatic milking system was associated with an increased risk of SARA5.8 (odds ratio: 10) and SARA6.0 (odds ratio: 11). The use of corn silage was associated with an increased risk of SARA5.8 (odds ratio: 21), whereas the use of monensin was associated with a decreased risk of SARA5.8 (odds ratio: 0.02). Our study shows that rpH is quite variable among farms, but also among animals on the same farm. We also show that multiple animal and farm characteristics are associated with rpH variability and the risk of SARA under commercial conditions.

Detecting gene signature activation in breast cancer in an absolute, single-patient manner.

BACKGROUND: The ability to reliably identify the state (activated, repressed, or latent) of any molecular process in the tumor of a patient from an individual whole-genome gene expression profile obtained from microarray or RNA sequencing (RNA-seq) promises important clinical utility. Unfortunately, all previous bioinformatics tools are only applicable in large and diverse panels of patients, or are limited to a single specific pathway/process (e.g. proliferation). METHODS: Using a panel of 4510 whole-genome gene expression profiles from 10 different studies we built and selected models predicting the activation status of a compendium of 1733 different biological processes. Using a second independent validation dataset of 742 patients we validated the final list of 1773 models to be included in a de novo tool entitled absolute inference of patient signatures (AIPS). We also evaluated the prognostic significance of the 1773 individual models to predict outcome in all and in specific breast cancer subtypes. RESULTS: We described the development of the de novo tool entitled AIPS that can identify the activation status of a panel of 1733 different biological processes from an individual breast cancer microarray or RNA-seq profile without recourse to a broad cohort of patients. We demonstrated that AIPS is stable compared to previous tools, as the inferred pathway state is not affected by the composition of a dataset. We also showed that pathway states inferred by AIPS are in agreement with previous tools but use far fewer genes. We determined that several AIPS-defined pathways are prognostic across and within molecularly and clinically define subtypes (two-sided log-rank test false discovery rate (FDR) <5%). Interestingly, 74.5% (1291/1733) of the models are able to distinguish patients with luminal A cancer from those with luminal B cancer (Fisher's exact test FDR <5%). CONCLUSION: AIPS represents the first tool that would allow an individual breast cancer patient to obtain a thorough knowledge of the molecular processes active in their tumor from only one individual gene expression (N-of-1) profile.