NobleBlocks

Research index

hiPRS algorithm process flow.

(14599915), (14599918), (9449680) and 7 others

Figshare10 Feb 2023Open access

NobleIDark:/48914/w1/20261005/4F58A597DOI10.1371/journal.pone.0281618.g002

Abstract

<p><b>(A)</b> Input data is a list of genotype-level SNPs. <b>(B)</b> Focusing on the positive class only, the algorithm exploits FIM (<i>apriori</i> algorithm) to build a list of candidate interactions of any desired order, retaining those that have an empirical frequency above a given threshold <i>δ</i>. This leads to a filtered set of terms in the form of sequences of pairs of SNP and associated categorical level (i.e., allele frequency in this example). The sequences can include from a single SNP-allele pair up to a maximum number of pairs defined by the user (<i>l</i><sub>max</sub>). <b>(C)</b> The whole training data is then scanned, searching for these sequences and deriving a re-encoded dataset where interaction terms are binary features (i.e., 1 if sequence <i>i</i> is observed in <i>j</i>-th patient genotype, 0 otherwise). From this dataset we can compute the MI between each interaction and the outcome and <b>(D)</b> obtain a ranked list (<i>I</i><sub><i>δ</i></sub>) based on this metric. <b>(E)</b> Starting from the interaction at the top of <i>I</i><sub><i>δ</i></sub>, <i>hi</i>PRS constructs <i>I</i><sub><i>K</i></sub>, selecting <i>K</i> (where <i>K</i> is user-specified) terms through the greedy optimization of the ratio between MI (<i>relevance</i>) and a suitable measure of similarity for interactions (<i>redundancy)</i> (cf. Algorithm 1, <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0281618#sec013" target="_blank">Materials and methods</a>). This leads to a set of predictive, yet diverse, interactions that <b>(F)</b> we use to define the score weighting their contribution by fitting a LR model and retaining the corresponding <i>β</i> coefficients.</p>

Discussion

Topics

More

Suggest a correction or report