Last change: 04.March.2021

# This directory contains files used for the analyses.

Existing files:

# input matrices
Each input matrix was combined by PolB counts matrix of 0.22-1.6/3.0 um metagenomic samples and each  of the V9 metabarcoding matrices (corresponding to different size fractions). Read counts in the newly generated matrices were normalized using centered log-ratio (clr) transformation after adding a pseudo count of one to all matrix elements because zero cannot be transformed in clr. Following clr normalization, we filtered out low-abundance OTUs with a lower quartile (Q1) filtering approach. Specifically, OTUs were retained in the matrices when their clr-normalized abundance was higher than Q1 (among the non-zero counts in the original count matrix prior to the addition of a pseudo count of one) in at least five samples. Normalization and filtering were separately applied to polB and V9. 

processed_input_matrix_size_0.8-inf.tsv
processed_input_matrix_size_0.8-5.tsv
processed_input_matrix_size_5-20.tsv
processed_input_matrix_size_20-180.tsv
processed_input_matrix_size_180-2000.tsv 

# edgelist
We kept one association with the highest absolute weight among all positive and negative associations between a given polB OTU and eukaryotic lineage for the assessment.

associations_used_in_assessment.edgelist

Reference:
Meng L, Endo H, Blanc-Mathieu R, Chaffron S, Hernandez-Velazquez R, Kaneko H, Ogata H. 2020. Quantitative assessment of NCLDV-host interactions predicted by co-occurrence analyses. bioRxiv doi.org/10.1101/2020.10.16.342030

If any question please contact: lingjie@kuicr.kyoto-u.ac.jp