Domain Adaptation Principal Component Analysis: Base Linear Method for Learning with Out-of-Distribution Data

CODI: Enhancing machine learning-based molecular profiling through contextual out-of-distribution integration

Molecular analytics increasingly utilize machine learning (ML) for predictive modeling based on data acquired through molecular profiling technologies. However, developing robust models that accurately capture physiological phenotypes is challenged by the dynamics inherent to biological systems, variability stemming from analytical procedures, and the resource-intensive nature of obtaining sufficiently representative datasets. Here, we propose and evaluate a new method: Contextual Out-of-Distribution Integration (CODI). Based on experimental observations, CODI generates synthetic data that integrate unrepresented sources of variation encountered in real-world applications into a given molecular fingerprint dataset. By augmenting a dataset with out-of-distribution variance, CODI enables an ML model to better generalize to samples beyond the seed training data, reducing the need for extensive experimental data collection. Using three independent longitudinal clinical studies and a case-control study, we demonstrate CODI's application to several classification tasks involving vibrational spectroscopy of human blood. We showcase our approach's ability to enable personalized fingerprinting for multiyear longitudinal molecular monitoring and enhance the robustness of trained ML models for improved disease detection. Our comparative analyses reveal that incorporating CODI into the classification workflow consistently leads to increased robustness against data variability and improved predictive accuracy.

Susceptibility of AutoML mortality prediction algorithms to model drift caused by the COVID pandemic

BACKGROUND

Concept drift and covariate shift lead to a degradation of machine learning (ML) models. The objective of our study was to characterize sudden data drift as caused by the COVID pandemic. Furthermore, we investigated the suitability of certain methods in model training to prevent model degradation caused by data drift.

METHODS

We trained different ML models with the H2O AutoML method on a dataset comprising 102,666 cases of surgical patients collected in the years 2014-2019 to predict postoperative mortality using preoperatively available data. Models applied were Generalized Linear Model with regularization, Default Random Forest, Gradient Boosting Machine, eXtreme Gradient Boosting, Deep Learning and Stacked Ensembles comprising all base models. Further, we modified the original models by applying three different methods when training on the original pre-pandemic dataset: (Rahmani K, et al, Int J Med Inform 173:104930, 2023) we weighted older data weaker, (Morger A, et al, Sci Rep 12:7244, 2022) used only the most recent data for model training and (Dilmegani C, 2023) performed a z-transformation of the numerical input parameters. Afterwards, we tested model performance on a pre-pandemic and an in-pandemic data set not used in the training process, and analysed common features.

RESULTS

The models produced showed excellent areas under receiver-operating characteristic and acceptable precision-recall curves when tested on a dataset from January-March 2020, but significant degradation when tested on a dataset collected in the first wave of the COVID pandemic from April-May 2020. When comparing the probability distributions of the input parameters, significant differences between pre-pandemic and in-pandemic data were found. The endpoint of our models, in-hospital mortality after surgery, did not differ significantly between pre- and in-pandemic data and was about 1% in each case. However, the models varied considerably in the composition of their input parameters. None of our applied modifications prevented a loss of performance, although very different models emerged from it, using a large variety of parameters.

CONCLUSIONS

Our results show that none of our tested easy-to-implement measures in model training can prevent deterioration in the case of sudden external events. Therefore, we conclude that, in the presence of concept drift and covariate shift, close monitoring and critical review of model predictions are necessary.

Transmorph: a unifying computational framework for modular single-cell RNA-seq data integration

Data integration of single-cell RNA-seq (scRNA-seq) data describes the task of embedding datasets gathered from different sources or experiments into a common representation so that cells with similar types or states are embedded close to one another independently from their dataset of origin. Data integration is a crucial step in most scRNA-seq data analysis pipelines involving multiple batches. It improves data visualization, batch effect reduction, clustering, label transfer, and cell type inference. Many data integration tools have been proposed during the last decade, but a surge in the number of these methods has made it difficult to pick one for a given use case. Furthermore, these tools are provided as rigid pieces of software, making it hard to adapt them to various specific scenarios. In order to address both of these issues at once, we introduce the transmorph framework. It allows the user to engineer powerful data integration pipelines and is supported by a rich software ecosystem. We demonstrate transmorph usefulness by solving a variety of practical challenges on scRNA-seq datasets including joint datasets embedding, gene space integration, and transfer of cycle phase annotations. transmorph is provided as an open source python package.

Omics data integration in computational biology viewed through the prism of machine learning paradigms

Important quantities of biological data can today be acquired to characterize cell types and states, from various sources and using a wide diversity of methods, providing scientists with more and more information to answer challenging biological questions. Unfortunately, working with this amount of data comes at the price of ever-increasing data complexity. This is caused by the multiplication of data types and batch effects, which hinders the joint usage of all available data within common analyses. Data integration describes a set of tasks geared towards embedding several datasets of different origins or modalities into a joint representation that can then be used to carry out downstream analyses. In the last decade, dozens of methods have been proposed to tackle the different facets of the data integration problem, relying on various paradigms. This review introduces the most common data types encountered in computational biology and provides systematic definitions of the data integration problems. We then present how machine learning innovations were leveraged to build effective data integration algorithms, that are widely used today by computational biologists. We discuss the current state of data integration and important pitfalls to consider when working with data integration tools. We eventually detail a set of challenges the field will have to overcome in the coming years.