\@writefile{lof}{\contentsline {figure}{\numberline {3.1}{\ignorespaces \textbf {Number of peaks in GM12878} called by ENCODE for each TF ChIP-seq experiment. The different TFs are colored by type, as defined by \citep {cheng_understanding_2012} : sequence specific TF (TFSS), non specific TF (TFNS), chromatin structure (ChromStr), chromatin modifier (ChromModif), RNAPII associated factors (Pol2), RNAPIII associated factors (Pol3) and others. The horizontal dashed lines indicate 20'000 and 40'000.\relax }}{32}{figure.caption.20}}
\newlabel{encode_peaks_gm12878_peak_number}{{3.1}{32}{\textbf {Number of peaks in GM12878} called by ENCODE for each TF ChIP-seq experiment. The different TFs are colored by type, as defined by \citep {cheng_understanding_2012} : sequence specific TF (TFSS), non specific TF (TFNS), chromatin structure (ChromStr), chromatin modifier (ChromModif), RNAPII associated factors (Pol2), RNAPIII associated factors (Pol3) and others. The horizontal dashed lines indicate 20'000 and 40'000.\relax }{figure.caption.20}{}}
\@writefile{lof}{\contentsline {figure}{\numberline {3.2}{\ignorespaces \textbf {Proportion of peaks with a motif in GM12878}, for each TF ChIP-seq experiment, in green. Assuming that a TF binds to DNA through its motif, the motif should be nearby the peak center. Thus the center of each peak was scanned using a PWM describing the TF binding specificity. Each TF was associated to a log-odd PWM contained either in JASPAR Core vertebrate 2014 \citep {mathelier_jaspar_2014}, HOCOMOCO v10 \citep {kulakovskiy_hocomoco:_2016} or Jolma \citep {jolma_dna-binding_2013} collection. If a motif instance with a score corresponding to a pvalue higher or equal to $1\cdot 10^{-4}$ could be found, the peak was considered bearing a motif. The different TFs are colored by type, as defined by \citep {cheng_understanding_2012} : sequence specific TF (TFSS), non specific TF (TFNS), chromatin structure (ChromStr), chromatin modifier (ChromModif), RNAPII associated factors (Pol2), RNAPIII associated factors (Pol3) and others. The horizontal dashed line indicates 0.5.\relax }}{32}{figure.caption.21}}
\newlabel{encode_peaks_gm12878_motif_prop}{{3.2}{32}{\textbf {Proportion of peaks with a motif in GM12878}, for each TF ChIP-seq experiment, in green. Assuming that a TF binds to DNA through its motif, the motif should be nearby the peak center. Thus the center of each peak was scanned using a PWM describing the TF binding specificity. Each TF was associated to a log-odd PWM contained either in JASPAR Core vertebrate 2014 \citep {mathelier_jaspar_2014}, HOCOMOCO v10 \citep {kulakovskiy_hocomoco:_2016} or Jolma \citep {jolma_dna-binding_2013} collection. If a motif instance with a score corresponding to a pvalue higher or equal to $1\cdot 10^{-4}$ could be found, the peak was considered bearing a motif. The different TFs are colored by type, as defined by \citep {cheng_understanding_2012} : sequence specific TF (TFSS), non specific TF (TFNS), chromatin structure (ChromStr), chromatin modifier (ChromModif), RNAPII associated factors (Pol2), RNAPIII associated factors (Pol3) and others. The horizontal dashed line indicates 0.5.\relax }{figure.caption.21}{}}
\@writefile{toc}{\contentsline {section}{\numberline {3.3}Nucleosome organization around transcription factor binding sites}{35}{section.3.3}}
\@writefile{lof}{\contentsline {figure}{\numberline {3.3}{\ignorespaces \textbf {Chromatin pattern around TF binding sites in GM12878 :} \textbf {A} For each peaklist, nucleosome occupancy was measured +/- 1kb around each individual TFBS using 10bp bins. The TFBS were then classified into 4 classes according to their nucleosome patterns using a ChIPPartitioning, allowing the patterns to be flipped and shifted. Each TF binding site was assigned a probability to belong to each of the 4 classes with a given values of shift and flip. To assess the extent of a given TF to i) display nucleosomes arrays on its flank and ii) to have nucleosome positioned with respect to its binding sites, array density and shift probability standard deviation have been measured for each class. Classes having a mean array density above 0.4 and a shift probability standard deviation under 3.5 and other custom classes are highlighted. Classes are named using the TF, the laboratory which produced the data and the class number (from 1 to 4). \textbf {B} Examples of class patterns corresponding to some of the highlighted classes for CTCF, ATF3, YY1, EBF1 and ZNF143. MNase profiles (red) were allowed to be shifted and flipped and DNaseI (blue), TSS density (violet) and sequence conservation (green) were overlaid according to MNase classification (taking into account both shift and flip). The y-axis scale represent the proportion of the highest signal for each chromatin pattern.\relax }}{36}{figure.caption.22}}
\newlabel{encode_peaks_array_measure}{{3.3}{36}{\textbf {Chromatin pattern around TF binding sites in GM12878 :} \textbf {A} For each peaklist, nucleosome occupancy was measured +/- 1kb around each individual TFBS using 10bp bins. The TFBS were then classified into 4 classes according to their nucleosome patterns using a ChIPPartitioning, allowing the patterns to be flipped and shifted. Each TF binding site was assigned a probability to belong to each of the 4 classes with a given values of shift and flip. To assess the extent of a given TF to i) display nucleosomes arrays on its flank and ii) to have nucleosome positioned with respect to its binding sites, array density and shift probability standard deviation have been measured for each class. Classes having a mean array density above 0.4 and a shift probability standard deviation under 3.5 and other custom classes are highlighted. Classes are named using the TF, the laboratory which produced the data and the class number (from 1 to 4). \textbf {B} Examples of class patterns corresponding to some of the highlighted classes for CTCF, ATF3, YY1, EBF1 and ZNF143. MNase profiles (red) were allowed to be shifted and flipped and DNaseI (blue), TSS density (violet) and sequence conservation (green) were overlaid according to MNase classification (taking into account both shift and flip). The y-axis scale represent the proportion of the highest signal for each chromatin pattern.\relax }{figure.caption.22}{}}
\@writefile{toc}{\contentsline {section}{\numberline {3.4}The case of CTCF, RAD21, SMC3, YY1 and ZNF143}{37}{section.3.4}}
\newlabel{encode_peaks_section_ctcf_rad21_smc3_yy1_znf143}{{3.4}{37}{The case of CTCF, RAD21, SMC3, YY1 and ZNF143}{section.3.4}{}}
\@writefile{lof}{\contentsline {figure}{\numberline {3.4}{\ignorespaces \textbf { Colocalization with CTCF peaks in GM12878 cells : } \textbf {A} Proportion of peaks for different TFs having a CTCF peak within 10bp, 50bp and 100bp. The colours indicate different TFs. The CTCF peaklist used as reference to assess CTCF presence was CTCF.Sydh (in red), the two RAD21 peaklists are RAD21.Haib and RAD21.Sydh respectively (in blue), the SMC3 peaklist is SMC3.Sydh (in green), the YY1 peaklist is YY1.Haib (in orange) and the ZNF143 peaklist is ZNF143.Sydh (in violet). \textbf {B} Venn diagrams showing the proportion of peaks for each TF with i) an instance of its own motif, ii) a CTCF.Sydh peak within 100bp, iii) both or iv) neither of them. RAD21 and SMC3 are not represented as there is no PWM available to describe their sequence specificity. \textbf {C} ChIPPartitioning classification with shift and flip of MNase patterns +/- 1kb of YY1.Haib peaks using 10bp bins. YY1 peaks with (upper row) and without (lower row) a CTCF peak within 100bp. Two classes were used to account for "typical" and "non-typical" looking MNase patterns. DNaseI (blue), TSS density (violet) and sequence conservation (green) were overlaid according to MNase classification (taking into account both shift and flip). The number at the upper right corner of each plot indicate the overall class probability. The number of YY1 peaks is slightly smaller than in B) because peaks showing no MNase reads were not included in the classification analysis. Peaklists are named using the TF together with the laboratory which produced the data.\relax }}{38}{figure.caption.23}}
\newlabel{encode_peaks_colocalization_ctcf}{{3.4}{38}{\textbf { Colocalization with CTCF peaks in GM12878 cells : } \textbf {A} Proportion of peaks for different TFs having a CTCF peak within 10bp, 50bp and 100bp. The colours indicate different TFs. The CTCF peaklist used as reference to assess CTCF presence was CTCF.Sydh (in red), the two RAD21 peaklists are RAD21.Haib and RAD21.Sydh respectively (in blue), the SMC3 peaklist is SMC3.Sydh (in green), the YY1 peaklist is YY1.Haib (in orange) and the ZNF143 peaklist is ZNF143.Sydh (in violet). \textbf {B} Venn diagrams showing the proportion of peaks for each TF with i) an instance of its own motif, ii) a CTCF.Sydh peak within 100bp, iii) both or iv) neither of them. RAD21 and SMC3 are not represented as there is no PWM available to describe their sequence specificity. \textbf {C} ChIPPartitioning classification with shift and flip of MNase patterns +/- 1kb of YY1.Haib peaks using 10bp bins. YY1 peaks with (upper row) and without (lower row) a CTCF peak within 100bp. Two classes were used to account for "typical" and "non-typical" looking MNase patterns. DNaseI (blue), TSS density (violet) and sequence conservation (green) were overlaid according to MNase classification (taking into account both shift and flip). The number at the upper right corner of each plot indicate the overall class probability. The number of YY1 peaks is slightly smaller than in B) because peaks showing no MNase reads were not included in the classification analysis. Peaklists are named using the TF together with the laboratory which produced the data.\relax }{figure.caption.23}{}}
\@writefile{lof}{\contentsline {figure}{\numberline {3.5}{\ignorespaces \textbf {Nucleosome free region at CTCF binding sites} \textbf {a} The length are represented as boxplots. The CTCF binding sites are divided into subgroups according to additional presence of SCM3, RAD21, YY1 or ZNF143. The number of binding sites in each subgroup is indicated in red above the boxplots. The presence of SMC3 only, RAD21 only and SMC3 and RAD21 together are indicated in violet, blue and orange respectively. \textbf {B} The proportion of peaks (in green), in each subgroup, having a TSS within a 1kb.\relax }}{39}{figure.caption.24}}
\newlabel{encode_peaks_ctcf_ndr}{{3.5}{39}{\textbf {Nucleosome free region at CTCF binding sites} \textbf {a} The length are represented as boxplots. The CTCF binding sites are divided into subgroups according to additional presence of SCM3, RAD21, YY1 or ZNF143. The number of binding sites in each subgroup is indicated in red above the boxplots. The presence of SMC3 only, RAD21 only and SMC3 and RAD21 together are indicated in violet, blue and orange respectively. \textbf {B} The proportion of peaks (in green), in each subgroup, having a TSS within a 1kb.\relax }{figure.caption.24}{}}
\@writefile{lof}{\contentsline {figure}{\numberline {3.6}{\ignorespaces \textbf {Possible interaction scenarios between TFs} \textbf {A} Indirect co-binding. The TFs dimerize and bind together on DNA. \textbf {B} Indirect co-binding. Both TF dimerize but only one binds the DNA, the other (the blue) is the tethering factor. \textbf {C} Independent co-binding. Both TF bind in close vicinity but without forming a complex. Both TFs may not be necessarily bound at the same time. \textbf {D} Interference. Both motifs partially or totally overlap each other. Whether only one TF or both can bind at the same time is unknown.\relax }}{41}{figure.caption.25}}
\newlabel{encode_peaks_tf_association}{{3.6}{41}{\textbf {Possible interaction scenarios between TFs} \textbf {A} Indirect co-binding. The TFs dimerize and bind together on DNA. \textbf {B} Indirect co-binding. Both TF dimerize but only one binds the DNA, the other (the blue) is the tethering factor. \textbf {C} Independent co-binding. Both TF bind in close vicinity but without forming a complex. Both TFs may not be necessarily bound at the same time. \textbf {D} Interference. Both motifs partially or totally overlap each other. Whether only one TF or both can bind at the same time is unknown.\relax }{figure.caption.25}{}}
\@writefile{toc}{\contentsline {section}{\numberline {3.5}CTCF and JunD interactomes}{41}{section.3.5}}
\@writefile{lof}{\contentsline {figure}{\numberline {3.7}{\ignorespaces \textbf {CTCF motif association} measured around the binding sites of different TFs. For a each TF, its binding sites, +/- 500bp, were searched for the presence of i) the TF motif and ii) CTCF motif. For each TF, a 2x2 contingency table was created with the number of peaks having i) both motifs, ii) the TF motif only, iii) CTCF motif only and iv) no motif. \textbf {A} Odd ratio (OR) of the exact Fisher test performed on each TF contingency table. The ORs are displayed with their 95\% confidence interval (CI). ORs > 1 - that is, with 1 not part of the 95\%CI - are labeled in green and indicate an association of both motifs more frequent than expected by chance. ORs < 1 are labeled in red and indicate a repulsion of both motifs more frequence than expected by chance. The CTCF dataset ORs are too high to be represented in this plot. \textbf {B} Density of CTCF motif occurrence at the absolute distance of different TF binding sites (peak centers) which also have their own motif present (at distance 0). The rows were standardized and aggregated using the Euclidean distance. \textbf {C} Same as in (B) but for TF binding sites that does not have their own motif. The absence of CTCF motif within the first 70bp around CTCF binding sites is explained by the peak processing (see section \ref {encode_peaks_methods_data}).\relax }}{42}{figure.caption.26}}
\newlabel{encode_peaks_ctcf_association}{{3.7}{42}{\textbf {CTCF motif association} measured around the binding sites of different TFs. For a each TF, its binding sites, +/- 500bp, were searched for the presence of i) the TF motif and ii) CTCF motif. For each TF, a 2x2 contingency table was created with the number of peaks having i) both motifs, ii) the TF motif only, iii) CTCF motif only and iv) no motif. \textbf {A} Odd ratio (OR) of the exact Fisher test performed on each TF contingency table. The ORs are displayed with their 95\% confidence interval (CI). ORs > 1 - that is, with 1 not part of the 95\%CI - are labeled in green and indicate an association of both motifs more frequent than expected by chance. ORs < 1 are labeled in red and indicate a repulsion of both motifs more frequence than expected by chance. The CTCF dataset ORs are too high to be represented in this plot. \textbf {B} Density of CTCF motif occurrence at the absolute distance of different TF binding sites (peak centers) which also have their own motif present (at distance 0). The rows were standardized and aggregated using the Euclidean distance. \textbf {C} Same as in (B) but for TF binding sites that does not have their own motif. The absence of CTCF motif within the first 70bp around CTCF binding sites is explained by the peak processing (see section \ref {encode_peaks_methods_data}).\relax }{figure.caption.26}{}}
\@writefile{lot}{\contentsline {table}{\numberline {3.1}{\ignorespaces \textbf {Identified associations : } Details of all the TF associations identified, as well as the possible molecular mechanisms explaining them. The columns 'TF${_A}$' and 'TF${_B}$' refer to the TF involved in the association, 'Motif.ass.' to whether both motif are associated together ('positive') or repel each other ('negative'), as measured by the Fisher test, 'Type' to the proposed interaction mechanism between both TFs, 'Binder' to the TF binding DNA in case of an indirect co-binding, the value 'both' means that both tethering complexes may exist, 'Reported' to whether this interaction has already been reported in one of the following study \cite {wang_sequence_2012, neph_expansive_2012, consortium_integrated_2012, guo_high_2012} and 'Validated' to whether this physical association is experimentally validated and reported in BioGRID v.3.4.145 \citep {chatr-aryamontri_biogrid_2017}.\relax }}{43}{table.caption.27}}
\newlabel{encode_peaks_association_table}{{3.1}{43}{\textbf {Identified associations : } Details of all the TF associations identified, as well as the possible molecular mechanisms explaining them. The columns 'TF${_A}$' and 'TF${_B}$' refer to the TF involved in the association, 'Motif.ass.' to whether both motif are associated together ('positive') or repel each other ('negative'), as measured by the Fisher test, 'Type' to the proposed interaction mechanism between both TFs, 'Binder' to the TF binding DNA in case of an indirect co-binding, the value 'both' means that both tethering complexes may exist, 'Reported' to whether this interaction has already been reported in one of the following study \cite {wang_sequence_2012, neph_expansive_2012, consortium_integrated_2012, guo_high_2012} and 'Validated' to whether this physical association is experimentally validated and reported in BioGRID v.3.4.145 \citep {chatr-aryamontri_biogrid_2017}.\relax }{table.caption.27}{}}
\@writefile{lof}{\contentsline {figure}{\numberline {3.8}{\ignorespaces \textbf {EBF1 binding sites} stand on the edge of a nucleosome. \textbf {A} Nucleosome dyad distributions around the EBF1 binding sites (from the Haib dataset). The dyad distributions have been measured from two independent datasets : i) MNase-seq data released by the ENCODE Consortium (in red) and by Gaffney et al. (in blue) \citep {gaffney_controls_2012}. \textbf {B} Dinucleotide frequencies around the dyads of the nucleosomes that have an EBF1 binding site within 100bp. \textbf {C} Motif frequency around the dyads of nucleosomes that have an EBF1 binding site within 100bp. The abrupt decrease of EBF1 motif frequency at +/- 100bp reflects the nucleosome selection process.\relax }}{47}{figure.caption.28}}
\newlabel{encode_peaks_ebf1}{{3.8}{47}{\textbf {EBF1 binding sites} stand on the edge of a nucleosome. \textbf {A} Nucleosome dyad distributions around the EBF1 binding sites (from the Haib dataset). The dyad distributions have been measured from two independent datasets : i) MNase-seq data released by the ENCODE Consortium (in red) and by Gaffney et al. (in blue) \citep {gaffney_controls_2012}. \textbf {B} Dinucleotide frequencies around the dyads of the nucleosomes that have an EBF1 binding site within 100bp. \textbf {C} Motif frequency around the dyads of nucleosomes that have an EBF1 binding site within 100bp. The abrupt decrease of EBF1 motif frequency at +/- 100bp reflects the nucleosome selection process.\relax }{figure.caption.28}{}}
\@writefile{loa}{\contentsline {algocf}{\numberline {1}{\ignorespaces Searches the coordinates of the NDR using the posterior nucleosome and nucleosome free class probabilities, for a region $R_i$, from its central position.\relax }}{53}{algocf.1}}
\citation{mathelier_jaspar_2014}
\citation{kulakovskiy_hocomoco:_2016}
\citation{ambrosini_pwmscan:_2018}
\citation{ambrosini_chip-seq_2016-1}
\citation{ambrosini_chip-seq_2016-1}
\citation{ambrosini_chip-seq_2016-1}
\citation{gaffney_controls_2012}
\@writefile{toc}{\contentsline {subsection}{\numberline {3.7.6}CTCF and JunD interactors}{54}{subsection.3.7.6}}
\citation{ambrosini_signal_2003}
\@writefile{toc}{\contentsline {subsection}{\numberline {3.7.7}EBF1 and nucleosome}{55}{subsection.3.7.7}}