arXiv subjects

Richard Tjörnhammar

Publications and source records attributed to Richard Tjörnhammar.

3 recordsLinked to original sources

arXiv2024

Happy and Immersive Clustering Segmentations of Biological Co-Expression Patterns

In this work, we present an approach for evaluating segmentation strategies and solving the biological problem of creating robust interpretable maps of biological data by employing wards agglomerative hierarchical clustering applied to coexpression coordinates to deduce a faithful representation of the input. We adopt and quantify two analyte-centric metrics named happiness and immersiveness, one for describing the suitability of a single analyte concerning the segmentation as well as a second metric for describing how well the segmentation catches the underlying data variation. We show that these two functions drive aggregation and segregation of segmentation respectively and can produce trustworthy segmentation solutions. We discover that the immersiveness metric exhibits higher-order phase transition properties in its derivative to cluster numbers. Finally, we find that the cluster representations and label annotations, in the case with clusters of high immersiveness, correspond to compositionally inferred labels with the highest specificity. The interconnectedness mirrors the potential relationships between cluster representations, label annotations, and inferred labels, emphasizing the intricate nature of biology and the representation of the specific expressions of gene products.

q-bio.QM↗

arXiv2022

Clustering Optimisation Method for Highly Connected Biological Data

Currently, data-driven discovery in biological sciences resides in finding segmentation strategies in multivariate data that produce sensible descriptions of the data. Clustering is but one of several approaches and sometimes falls short because of difficulties in assessing reasonable cutoffs, the number of clusters that need to be formed or that an approach fails to preserve topological properties of the original system in its clustered form. In this work, we show how a simple metric for connectivity clustering evaluation leads to an optimised segmentation of biological data. The novelty of the work resides in the creation of a simple optimisation method for clustering crowded data. The resulting clustering approach only relies on metrics derived from the inherent properties of the clustering. The new method facilitates knowledge for optimised clustering, which is easy to implement. We discuss how the clustering optimisation strategy corresponds to the viable information content yielded by the final segmentation. We further elaborate on how the clustering results, in the optimal solution, corresponds to prior knowledge of three different data sets.

q-bio.QM↗

arXiv2021

Exploratory Projection to Latent Structure Models for use in Transcriptomic Analysis

In this paper, we ask if it is possible to increase the interpretability in multivariate analysis by aligning and projecting covariates onto comparative subspaces. We demonstrate our method as well as the interpretative power of PLS decomposed models and how robust interpretability can lead to quantitative insights. We discuss the statistical properties of the PLS weights, $p$-values associated with specific axes, as well as their alignment properties. The applicability of this approach within life science is also demonstrated by applying it to three use cases of publically available datasets. Further we present hierarchical pathway enrichment results stemming from aligned $p$-values, which are compared with results derived from enrichment analysis, as an external validation of our method. We find that the method can uncover known results from genomics for all of the studied use cases, i.e. microarray data from multiple sclerosis and diabetes patients as well as RNA sequencing data from breast cancer patients.

q-bio.QM↗