Rosane Minghim

The identification of the site of origin, stage and driver genes of cancer is important for the improved prognosis of cancer patients. Publicly available somatic mutation data and machine learning techniques provide a framework for cancer stage classification. However, due to the heterogeneity and sparsity of somatic mutation data, classifying cancer stages is challenging. Researchers are keen to employ Artificial Intelligence (AI), and specifically Machine Learning (ML) techniques to detect cancer in the early stages. Most of the models provided, however, are very specific to particular cancer types and do not generalise. We propose a pipeline named Clustering with Latent Autoencoder Representations and Interpretable Technology (CLARITY) for somatic mutation analysis, which works in two ways. First, we developed an autoencoder-based clustering algorithm for desparsified somatic mutation data to identify cancer stages based on cancer hallmark scores. Second, we created a Deep Neural Network (DNN) that is tightly integrated with autoencoder-based representation learning and explainable AI (XAI) to classify these stages and interpret the hallmarks that influence cancer progression. These hallmarks are candidates for biomarkers.We demonstrate CLARITY on Breast Invasive Carcinoma (BRCA), achieving 99.47% accuracy and outperforming state-of-the-art methods. Furthermore, we show the robustness of our performance using statistical analysis, traditional ML algorithms and data visualisation techniques. This demonstrates how well our clustering algorithm can identify cancer subtypes and how our classifier handles sparse and noisy somatic mutation data.

Cancer is one of the most deadly diseases, costing millions of lives and billions of USD every year. There are different ways to identify the biomarkers that can be used to detect cancer types and subtypes. RNA sequencing is steadily taking the lead as the method of choice due to its ability to access global gene expression in biological samples and facilitate more flexible methods and robust analyses. Numerous studies have employed artificial intelligence (AI) and specifically machine learning techniques to detect cancer in its early stages. However, most of the models provided are very specific to particular cancer types and do not generalize. This paper proposes a deep learning and explainable AI (XAI) combined approach to classifying cancer subtypes and a deep learning-based approach for the classification of cancer types using BARRA:CuRDa, an RNA-seq database with 17 datasets for seven cancer types. One architecture is designed to classify cancer subtypes with around 100% accuracy, precision, recall, F1 score, and G-Mean. This architecture outperforms the previous methodologies for all individual datasets. The second architecture is designed to classify multiple cancer types; it classifies eight types within the neighborhood of 87% of validation accuracy, precision, recall, F1 score, and G-Mean. Within the same process, we employ XAI, which identifies 99 genes out of 58,735 input genes that could be potential biomarkers for different cancer types. We also perform Pathway Enrichment Analysis and Visual Analysis to establish the significance and robustness of our methodology. The proposed methodology can classify cancer types and subtypes with robust results and can be extended to other cancer types.

Large language models (LLMs) have recently made significant advances, excelling in tasks like question answering, summarization, and machine translation. However, their enormous size and hardware requirements make them less accessible to many in the machine learning community. To address this, techniques such as quantization, prefix tuning, weak supervision, low-rank adaptation, and prompting have been developed to customize these models for specific applications. While these methods have mainly improved text generation, their implications for the text classification task are not thoroughly studied. Our research intends to bridge this gap by investigating how variations like model size, pre-training objectives, quantization, low-rank adaptation, prompting, and various hyperparameters influence text classification tasks. Our overall conclusions show the following: 1—even with synthetic labels, fine-tuning works better than prompting techniques, and increasing model size does not always improve classification performance; 2—discriminatively trained models generally perform better than generatively pre-trained models; and 3—fine-tuning models at 16-bit precision works much better than using 8-bit or 4-bit models, but the performance drop from 8-bit to 4-bit is smaller than from 16-bit to 8-bit. In another scale of our study, we conducted experiments with different settings for low-rank adaptation (LoRA) and quantization, finding that increasing LoRA dropout negatively affects classification performance. We did not find a clear link between the LoRA attention dimension (rank) and performance, observing only small differences between standard LoRA and its variants like rank-stabilized LoRA and weight-decomposed LoRA. Additional observations to support model setup for classification tasks are presented in our analyses.

The abundance of digital documents offers a valuable chance to gain insights into public opinion, social structure, and dynamics. However, the scale and volume of these digital collections makes manual analysis approaches extremely costly and not scalable. In this paper, we study the potential of using automated methods from natural language processing and machine learning, in particular weak supervision strategies, to understand how news influence decision making in society. Besides proposing a weak supervision solution for the task, which replaces manual labeling to a certain extent, we propose an improvement of a recently published economic index. This index is known as economic policy uncertainty (EPU) index and has been shown to correlate to indicators such as firm investment, employment, and excess market returns. In summary, in this paper, we present an automated data efficient approach based on weak supervision and deep learning (BERT + WS) for identification of news articles about economical uncertainty and adapt the calculation of EPU to the proposed strategy. Experimental results reveal that our approach (BERT + WS) improves over the baseline method centered in keyword search, which is currently used to construct the EPU index. The improvement is over 20 points in precision, reducing the false positive rate typical to the use of keywords.

Data visualisation helps understanding data represented by multiple variables, also called features, stored in a large matrix where individuals are stored in lines and variable values in columns. These data structures are frequently called multidimensional spaces. A large set of mathematical tools, named frequently as multidimensional projections, aim to map such large spaces into 'visual spaces', that is, to 2 or 3 dimensions, where the aspect of that space can be visualised. While the final product is intuitive in that proximity between points - or iconic representation of points - indicate similarity relationships in the original space, understanding the formulation of the projection methods many times escapes researchers. In this paper, we illustrate ways of employing the visual results of multidimensional projection algorithms to understand and fine-tune the parameters of their mathematical framework. Some of the common mathematical common to these approaches are Laplacian matrices, Euclidian distance, Cosine distance, and statistical methods such as Kullback-Leibler divergence, employed to fit probability distributions and reduce dimensions. Two of the relevant algorithms in the data visualisation field are t-distributed stochastic neighbourhood embedding (t-SNE) and Least-Square Projection (LSP). These algorithms can be used to understand several ranges of mathematical functions including their impact on datasets. In this article, mathematical parameters of underlying techniques such as Principal Component Analysis (PCA) behind t-SNE and mesh reconstruction methods behind LSP are adjusted to reflect the properties afforded by the mathematical formulation. The results, supported by illustrative methods of the processes of LSP and t-SNE, are meant to inspire students in understanding the mathematics behind such methods, in order to apply them in effective data analysis tasks in multiple applications.

Data visualisation helps understanding data represented by multiple variables, also called features, stored in a large matrix where individuals are stored in lines and variable values in columns. These data structures are frequently called multidimensional spaces.In this paper, we illustrate ways of employing the visual results of multidimensional projection algorithms to understand and fine-tune the parameters of their mathematical framework. Some of the common mathematical common to these approaches are Laplacian matrices, Euclidian distance, Cosine distance, and statistical methods such as Kullback-Leibler divergence, employed to fit probability distributions and reduce dimensions. Two of the relevant algorithms in the data visualisation field are t-distributed stochastic neighbourhood embedding (t-SNE) and Least-Square Projection (LSP). These algorithms can be used to understand several ranges of mathematical functions including their impact on datasets. In this article, mathematical parameters of underlying techniques such as Principal Component Analysis (PCA) behind t-SNE and mesh reconstruction methods behind LSP are adjusted to reflect the properties afforded by the mathematical formulation. The results, supported by illustrative methods of the processes of LSP and t-SNE, are meant to inspire students in understanding the mathematics behind such methods, in order to apply them in effective data analysis tasks in multiple applications.