Giovanni Parmigiani

Giovanni Parmigiani

In this article I propose an approach for defining replicability for prediction rules. Motivated by a recent NAS report, I start from the perspective that replicability is obtaining consistent results across studies suitable to address the same prediction question, each of which has obtained its own data. I then discuss concept and issues in defining key elements of this statement. I focus specifically on the meaning of "consistent results" in typical utilization contexts, and propose a multi-agent framework for defining replicability, in which agents are neither partners nor adversaries. I recover some of the prevalent practical approaches as special cases. I hope to provide guidance for a more systematic assessment of replicability in machine learning.

Cathy Shyr, Pragya Sur, Giovanni Parmigiani et al.

Cross-study replicability is a powerful model evaluation criterion that emphasizes generalizability of predictions. When training cross-study replicable prediction models, it is critical to decide between merging and treating the studies separately. We study boosting algorithms in the presence of potential heterogeneity in predictor-outcome relationships across studies and compare two multi-study learning strategies: 1) merging all the studies and training a single model, and 2) multi-study ensembling, which involves training a separate model on each study and ensembling the resulting predictions. In the regression setting, we provide theoretical guidelines based on an analytical transition point to determine whether it is more beneficial to merge or to ensemble for boosting with linear learners. In addition, we characterize a bias-variance decomposition of estimation error for boosting with component-wise linear learners. We verify the theoretical transition point result in simulation and illustrate how it can guide the decision on merging vs. ensembling in an application to breast cancer gene expression data.

Keith Barnatchez, Kevin P. Josey, Rachel C. Nethery et al.

Data-driven decision making frequently relies on predicting counterfactual outcomes. In practice, researchers commonly train counterfactual prediction models on a source dataset to inform decisions on a possibly separate target population. Conformal prediction has arisen as a popular method for producing assumption-lean prediction intervals for counterfactual outcomes that would arise under different treatment decisions in the target population of interest. However, existing methods require that every confounding factor of the treatment-outcome relationship used for training on the source data is additionally measured in the target population, risking miscoverage if important confounders are unmeasured in the target population. In this paper, we introduce a computationally efficient debiased machine learning framework that allows for valid prediction intervals when only a subset of confounders is measured in the target population, a common challenge referred to as runtime confounding. Grounded in semiparametric efficiency theory, we show the resulting prediction intervals achieve desired coverage rates with faster convergence compared to standard methods. Through numerous synthetic and semi-synthetic experiments, we demonstrate the utility of our proposed method.

Yujie Wu, Giovanni Parmigiani, Boyu Ren

Multi-source domain adaptation (DA) aims at leveraging information from more than one source domain to make predictions in a target domain, where different domains may have different data distributions. Most existing methods for multi-source DA focus on classification problems while there is only limited investigation in the regression settings. In this paper, we fill in this gap through a two-step procedure. First, we extend a flexible single-source DA algorithm for classification through outcome-coarsening to enable its application to regression problems. We then augment our single-source DA algorithm for regression with ensemble learning to achieve multi-source DA. We consider three learning paradigms in the ensemble algorithm, which combines linearly the target-adapted learners trained with each source domain: (i) a multi-source stacking algorithm to obtain the ensemble weights; (ii) a similarity-based weighting where the weights reflect the quality of DA of each target-adapted learner; and (iii) a combination of the stacking and similarity weights. We illustrate the performance of our algorithms with simulations and a data application where the goal is to predict High-density lipoprotein (HDL) cholesterol levels using gut microbiome. We observe a consistent improvement in prediction performance of our multi-source DA algorithm over the routinely used methods in all these scenarios.

Gabriel Loewinger, Rolando Acosta Nunez, Rahul Mazumder et al.

It is increasingly common to encounter prediction tasks in the biomedical sciences for which multiple datasets are available for model training. Common approaches such as pooling datasets and applying standard statistical learning methods can result in poor out-of-study prediction performance when datasets are heterogeneous. Theoretical and applied work has shown $\textit{multi-study ensembling}$ to be a viable alternative that leverages the variability across datasets in a manner that promotes model generalizability. Multi-study ensembling uses a two-stage $\textit{stacking}$ strategy which fits study-specific models and estimates ensemble weights separately. This approach ignores, however, the ensemble properties at the model-fitting stage, potentially resulting in a loss of efficiency. We therefore propose $\textit{optimal ensemble construction}$, an $\textit{all-in-one}$ approach to multi-study stacking whereby we jointly estimate ensemble weights as well as parameters associated with each study-specific model. We prove that limiting cases of our approach yield existing methods such as multi-study stacking and pooling datasets before model fitting. We propose an efficient block coordinate descent algorithm to optimize the proposed loss function. We compare our approach to standard methods by applying it to a multi-country COVID-19 dataset for baseline mortality prediction. We show that when little data is available for a country before the onset of the pandemic, leveraging data from other countries can substantially improve prediction accuracy. Importantly, our approach outperforms multi-study stacking and other standard methods in this application. We further characterize the method's performance in simulations. Our method remains competitive with or outperforms multi-study stacking and other earlier methods across a range of between-study heterogeneity levels.

Zoe Guan, Giovanni Parmigiani, Danielle Braun et al.

Family history is a major risk factor for many types of cancer. Mendelian risk prediction models translate family histories into cancer risk predictions based on knowledge of cancer susceptibility genes. These models are widely used in clinical practice to help identify high-risk individuals. Mendelian models leverage the entire family history, but they rely on many assumptions about cancer susceptibility genes that are either unrealistic or challenging to validate due to low mutation prevalence. Training more flexible models, such as neural networks, on large databases of pedigrees can potentially lead to accuracy gains. In this paper, we develop a framework to apply neural networks to family history data and investigate their ability to learn inherited susceptibility to cancer. While there is an extensive literature on neural networks and their state-of-the-art performance in many tasks, there is little work applying them to family history data. We propose adaptations of fully-connected neural networks and convolutional neural networks to pedigrees. In data simulated under Mendelian inheritance, we demonstrate that our proposed neural network models are able to achieve nearly optimal prediction performance. Moreover, when the observed family history includes misreported cancer diagnoses, neural networks are able to outperform the Mendelian BRCAPRO model embedding the correct inheritance laws. Using a large dataset of over 200,000 family histories, the Risk Service cohort, we train prediction models for future risk of breast cancer. We validate the models using data from the Cancer Genetics Network.

Maya Ramchandran, Rajarshi Mukherjee, Giovanni Parmigiani

Adapting machine learning algorithms to better handle the presence of clusters or batch effects within training datasets is important across a wide variety of biological applications. This article considers the effect of ensembling Random Forest learners trained on clusters within a dataset with heterogeneity in the distribution of the features. We find that constructing ensembles of forests trained on clusters determined by algorithms such as k-means results in significant improvements in accuracy and generalizability over the traditional Random Forest algorithm. We begin with a theoretical exploration of the benefits of our novel approach, denoted as the Cross-Cluster Weighted Forest, and subsequently empirically examine its robustness to various data-generating scenarios and outcome models. Furthermore, we explore the influence of the data partitioning and ensemble weighting strategies on the benefits of our method over the existing paradigm. Finally, we apply our approach to cancer molecular profiling and gene expression datasets that are naturally divisible into clusters and illustrate that our approach outperforms classic Random Forest.

Theodore Huang, Gregory Idos, Christine Hong et al.

Improving existing widely-adopted prediction models is often a more efficient and robust way towards progress than training new models from scratch. Existing models may (a) incorporate complex mechanistic knowledge, (b) leverage proprietary information and, (c) have surmounted barriers to adoption. Compared to model training, model improvement and modification receive little attention. In this paper we propose a general approach to model improvement: we combine gradient boosting with any previously developed model to improve model performance while retaining important existing characteristics. To exemplify, we consider the context of Mendelian models, which estimate the probability of carrying genetic mutations that confer susceptibility to disease by using family pedigrees and health histories of family members. Via simulations we show that integration of gradient boosting with an existing Mendelian model can produce an improved model that outperforms both that model and the model built using gradient boosting alone. We illustrate the approach on genetic testing data from the USC-Stanford Cancer Genetics Hereditary Cancer Panel (HCP) study.

Zhun Deng, Frances Ding, Cynthia Dwork et al.

We investigate the power of censoring techniques, first developed for learning {\em fair representations}, to address domain generalization. We examine {\em adversarial} censoring techniques for learning invariant representations from multiple "studies" (or domains), where each study is drawn according to a distribution on domains. The mapping is used at test time to classify instances from a new domain. In many contexts, such as medical forecasting, domain generalization from studies in populous areas (where data are plentiful), to geographically remote populations (for which no training data exist) provides fairness of a different flavor, not anticipated in previous work on algorithmic fairness. We study an adversarial loss function for $k$ domains and precisely characterize its limiting behavior as $k$ grows, formalizing and proving the intuition, backed by experiments, that observing data from a larger number of domains helps. The limiting results are accompanied by non-asymptotic learning-theoretic bounds. Furthermore, we obtain sufficient conditions for good worst-case prediction performance of our algorithm on previously unseen domains. Finally, we decompose our mappings into two components and provide a complete characterization of invariance in terms of this decomposition. To our knowledge, our results provide the first formal guarantees of these kinds for adversarial invariant domain generalization.

Zoe Guan, Giovanni Parmigiani, Prasad Patil

A critical decision point when training predictors using multiple studies is whether studies should be combined or treated separately. We compare two multi-study prediction approaches in the presence of potential heterogeneity in predictor-outcome relationships across datasets: 1) merging all of the datasets and training a single learner, and 2) multi-study ensembling, which involves training a separate learner on each dataset and combining the predictions resulting from each learner. For ridge regression, we show analytically and confirm via simulation that merging yields lower prediction error than ensembling when the predictor-outcome relationships are relatively homogeneous across studies. However, as cross-study heterogeneity increases, there exists a transition point beyond which ensembling outperforms merging. We provide analytic expressions for the transition point in various scenarios, study asymptotic properties, and illustrate how transition point theory can be used for deciding when studies should be combined with an application from metagenomics.

Yujia Bao, Zhengyi Deng, Yan Wang et al.

PURPOSE: The medical literature relevant to germline genetics is growing exponentially. Clinicians need tools monitoring and prioritizing the literature to understand the clinical implications of the pathogenic genetic variants. We developed and evaluated two machine learning models to classify abstracts as relevant to the penetrance (risk of cancer for germline mutation carriers) or prevalence of germline genetic mutations. METHODS: We conducted literature searches in PubMed and retrieved paper titles and abstracts to create an annotated dataset for training and evaluating the two machine learning classification models. Our first model is a support vector machine (SVM) which learns a linear decision rule based on the bag-of-ngrams representation of each title and abstract. Our second model is a convolutional neural network (CNN) which learns a complex nonlinear decision rule based on the raw title and abstract. We evaluated the performance of the two models on the classification of papers as relevant to penetrance or prevalence. RESULTS: For penetrance classification, we annotated 3740 paper titles and abstracts and used 60% for training the model, 20% for tuning the model, and 20% for evaluating the model. The SVM model achieves 89.53% accuracy (percentage of papers that were correctly classified) while the CNN model achieves 88.95 % accuracy. For prevalence classification, we annotated 3753 paper titles and abstracts. The SVM model achieves 89.14% accuracy while the CNN model achieves 89.13 % accuracy. CONCLUSION: Our models achieve high accuracy in classifying abstracts as relevant to penetrance or prevalence. By facilitating literature review, this tool could help clinicians and researchers keep abreast of the burgeoning knowledge of gene-cancer associations and keep the knowledge bases for clinical decision support tools up to date.

Giovanni Parmigiani

The fuzzy ROC extends Receiver Operating Curve (ROC) visualization to the situation where some data points, falling in an indeterminacy region, are not classified. It addresses two challenges: definition of sensitivity and specificity bounds under indeterminacy; and visual summarization of the large number of possibilities arising from different choices of indeterminacy zones.