Unsupervised tensor decomposition-based method to extract candidate transcriptionfactors as histone modification bookmarks in post-mitotic transcriptional reactivation Y-h. Taguchi Department of Physics, Chuo University, Tokyo, Japan Turki Turki Department of Computer Science, King Abdulaziz University, Jeddah, Saudi Arabia bioRxiv https://www.biorxiv.org/content/10.1101/2020.09.23.309633v1

Introduction Transcription ← ⨉ → DNA replication ⇓ DNA replication ⇒ “Transcription → ⨉” ⇓ Genome state (TF binding, histone modification) → Reset? ⇓ No, should be bookmarked bookmarked ! ⇓ But how?

whole-genome histone modification profile GSE141081 x ijkms ∈ℝN ×2×4×3×2 ⇒ formatted as tensor i: genome regions of 25,000 bps j: cell lines 1:RPE1, 2:USO2 k: histone modification 1: H3K27ac, 2: H3K4me1, 3: H3K4me3, 4: input m:cell cycle 1:interphase, 2:prometaphase, 3 :anaphase/telophase s:replicates

Tensor decomposition (Tucker) by higher order singular value decomposition (HOSVD) x ijkms x ijkms ≃∑ l 1 =1 2 ∑ l 2 =1 4 ∑ l 3 =1 3 ∑ l 4 =1 2 ∑ l 5 =1 N G(l 1 l 2 l 3 l 4 l 5 )u l 1 j u l 2 k u l 3 m u l 4 s u l 5 i G u l 2 k u l 3 m ≃ ● ● ● ● ● ● u l 1 j

∑ l 1 =1 2 ∑ l 2 =1 4 ∑ l 3 =1 3 ∑ l 4 =1 2 ∑ l 5 =1 N G(l 1 l 2 l 3 l 4 l 5 )u l 1 j u l 2 k u l 3 m u l 4 s u l 5 i l 1 =1: no dependence upon cell lines 2 ≤ l 2 ≤4: some dependence upon histone modification l 3 =3: most significant reactivation during phases l 4 =1: no dependence upon replicates ∑ l 2 =2 4 G(1,l 2 ,3,1,l 5 )2 l l 5 5 =4 =4

Dependence upon cell cycle Reactivation Reactivation l l 3 3 =1 =1 l l 3 3 =2 =2 l l 3 3 =3 =3

control Histone modification dependence l l 2 2 =1 =1 l l 2 2 =2 =2 l l 2 2 =3 =3 l l 2 2 =4 =4

Cell line dependence Replicate dependence l l 1 1 =1 =1 l l 4 4 =1 =1

∑ l 2 =2 4 G(1,l 2 ,3,1,l 5 )2 l l 5 5 =4 =4

Attributing P-values to ith genomic region P i =P χ2 [> (u 4 i σ4 )2] Cumulative chi squared distribution Assuming that u 4i obeys Gaussian

507 DNA regions : BH criterion corrected P values less than 0.01 ⇓ 525 gene symbols are included

Are these genes associated with histone modification coincident with re-activation or bookmark?

Enrichment analysis TFs binding

Some biologically critical TFs are included.

Some biologically critical TFs are included.

Top 10 most frequently listed transcription factor (TF) families These are candidate bookmark TFs! These are candidate bookmark TFs!

Conclusion ● Bookmark mechanism of reactivation of transcription after DNA duplication was studied. ● TD based unsupervised FE was applied to three histone modification profiles before/after DNA replication ● Some candidate bookmark TFs were identified.