Readme for MCOT software package

General description

MCOT (Motifs Co-Occurrence Tool) is a software package for recognition of composite elements (CEs) in a single ChIP-seq dataset (Levitsky et al., 2019; Levitsky et al., 2020). CEs detected by MCOT include two potential binding sites of transcription factors (TFs) in all possible mutual orientations. MCOT considers CEs with a full/partial overlap of motifs or with a spacer in a certain range. Each potential CE recognized by MCOT contains the motif of immunoprecipitated TF in respective ChIP-seq experiment (anchor motif) and another motif (partner). Identical/distinct anchor and partner motifs imply the search for CEs of homotypic or eterotypic type (respectively). The web service WebMCOT Levitsky et al., 2022 implements MCOT software package.

Implementation

MCOT implemented in C++ and it can be conventionally compiled in Linux or Windows operating system. To run MCOT user should compile the corresponding source code file. Files mcot_anchor.cpp and mcot.cpp respect to one-partner and many partners options for Position Weight Matrix (PWM) model of a binding site. File anchor_pro.cpp respects to one-partner option, but it runs with arbitrary models of site, including not-PWM ones, e.g. BaMM (Siebert and Söding, 2016), and SiteGA (Tsukanov et al., 2022)

Installation

(Linux) Run in terminal (Packages “build-essential” and “cmake” should be installed on Ubuntu system):

cd <project>
mkdir tmp
cd tmp
cmake ..
make

All executable files for one-partner, many-partners and anchor-pro options will be in src/anchor_vs_one, src/anchor_vs_many and src/anchor_pro

(Windows) Run in terminal (Win -> Visual Studio 2017 -> Visual Studio Tools -> VC -> Native Tools x64. Else, you should install “CMake” module while VS 2017 installing)

cd <project>
mkdir tmp
cd tmp
cmake ..
MSBuild mcot-kernel.sln /p:Configuration=Release /p:Platform=Win32

Programs anchor_vs_many, anchor_vs_one and anchor_pro for one_partner, many_partners and arbitrary_models_one_partner options should be in src/anchor_vs_one_partner/Release/, src/anchor_vs_many_partners/Release/ and src/anchor_pro/Release/

Command line arguments

The command line for one-partner option:

./anchor_vs_one <1 fasta> <2 anchor.motif> <3 partner.motif> <4 minimal spacer length> <5 maximal spacer length> <6 path to whole-genome promoters> <7 pvalue_thr> <8 -log10[p-value]_thr> <9 asymmetry_fold(-log10(ERR))>

The command line for many-partner option:

./anchor_vs_many <1 fasta> <2 anchor.motif> <3 partners.library> <4 minimal spacer length> <5 maximal spacer length> <6 path to whole-genome promoters> <7 pvalue_thr> <8 -log10[p-value]_thr> <9 asymmetry_ratio(-log10(ERR))>

<1 fasta> = DNA sequences of ChIP-seq peaks in fasta format, a minimum recommended number of peaks is about 300-500, the maximum number is not restricted, however 5000-10000 or higher number of peaks requires a higher computation time than several thousands of peaks, hence about 1000-2000 peaks are enough. Sequences should have lengths substatially higher than lengths of recognition models for anchor and partner motifs to contain possible composite elememnts with an overlap or spacer.

<1 fasta> = DNA sequences of ChIP-seq peaks in fasta format, the minimum recommended number of peaks is about 200-300, the maximum number is not restricted, however 10000 or higher number of peaks requires higher computation time than several thousands of peaks. Sequences should have lengths substatially higher than lengths of recognition models for anchor and partner motifs to contain possible composite elememnts with an overlap or spacer.

<2 anchor.motif>, <3 partner.motif> = frequency matrices of the first and second motifs motifs in the standard format, e.g.

***

> Motif name
0.001    0.001    0.942    0.056
0.001    0.001    0.997    0.001
0.001    0.001    0.997    0.001
0.548    0.001    0.400    0.051
0.589    0.087    0.245    0.079
0.231    0.001    0.001    0.767
0.001    0.001    0.001    0.997
0.001    0.381    0.001    0.617
0.001    0.964    0.001    0.034
0.001    0.891    0.001    0.107
***

<3 partners.library> = for this parameter five options are available: “hs_core”, “mm_core”, “hs_full”, “mm_full” and “dapseq”. These values respect to the libraries of derived from ChIP-seq data Hocomoco core (396/353) and full (747/509) collections of motifs for human/murine TFs (Kulakovskiy et al., 2018); and the library of motifs derived from DAP-seq data Plant Cistrome collection of 514 motifs for A.thaliana TFs (O’Malley et al., 2016).

<4 minimal spacer length> = integer value from 0 to \ (the default value 0 is recommended, any positive value restricts short spacers)

<5 maximal spacer length> = integer value from 0 to 100 (the default value 29)

<6 path to whole-genome promoters> = a path to the whole-genome dataset of promoters, three folders “hs”, “mm” and “at” that imply application of H.sapiens, M.musculus and A.thaliana promoter datasets for setting of thresholds for input motifs.

<7 pvalue_thr> = recognition threshold of motifs transformed to the logarithmic -log10(ERR) scale of Expected Recognition Rate (ERR), ERR is computed as a recognition rate for the whole-genome set o promoters of protein-coding genes, default value 0.0005

<8 -log10[p-value]_thr> = threshold to display the significances of enrichment of CEs in output data (the default value 10)

<9 asymmetry_ratio(-log10(ERR))> = the ratio ratio_thr restricting ERR values of two motifs in asymmetrical CEs, e.g. for two value ERR1 and ERR2 for certain CE means ratio_thr = Max{-log10(ERR1), -log10(ERR2)} / Min{-log10(ERR1), -log10(ERR2)}, and ratio > ratio_thr and ratio < ratio_thr mean asymmetrical and symmetrical CEs, respectively.

The command line for anchor_pro option:

./anchor_pro <1 file_fasta> <2 motif1.profile> <3 motif2.profile> <4 int motif1.length> <5 int motif1.length> <6 int motif1.table_thr_err> <7 int motif1.table_thr_err> <8 int spacer_min> <9 int spacer_max> <10double pvalue_thr> <11double -log10[p-value]_thr> <12double asymmetry_ratio(-log10(ERR))>

<1 file_fasta> = DNA sequences of ChIP-seq peaks in fasta format

<2 motif1.profile> = Profile for the first model, see example profile of model 1

<3 motif2.profile> = Profile for the second model, see example profile of model 2

<4 int motif1.length> = integer value, length of the first model

<5 int motif1.length> = integer value, length of the second model

<6 int motif1.table_thr_err> = Table Threshold vs. ERR for the first motif, see example of distribution for model 1

<7 int motif2.table_thr_err> = Table Threshold vs. ERR for the second motif, see example of distribution for model 2

<8 int spacer_min> = integer value from 0 to \ (the default value 0 is recommended, any positive value restricts short spacers)

<9 int spacer_max> = integer value from 0 to 100 (the default value 29)

<10 pvalue_thr> = recognition threshold of motifs transformed to the logarithmic scale, -log10(ERR), default value 0.0005

<11 -log10[p-value]_thr> = threshold to display the significances of enrichment of CEs in output data (the default value 10)

<12 asymmetry_ratio(-log10(ERR))> = the ratio ratio_thr restricting ERR values of two motifs in asymmetrical CEs, e.g. for two value ERR1 and ERR2 for certain CE means ratio_thr = Max{-log10(ERR1), -log10(ERR2)} / Min{-log10(ERR1), -log10(ERR2)}, and ratio > ratio_thr and ratio < ratio_thr mean asymmetrical and symmetrical CEs, respectively.

Input data

MCOT requires (a) DNA sequences of ChIP-seq peaks and (b) anchor and partner motifs. We recommend application of a conventional de novo motif search tool, e.g. Homer (Heinz et al., 2010) and STREME (Bailey et al., 2021) to define an anchor motif.

MCOT have two options for definition of the partner motif:

• a matrix of partner motif (one-partner option);

• a library of known motifs (many partners option).

<anchor_pro> requires input files Table Threshold vs. ERR for both models. For a PWM model the respictive file can be taken as the output files of runs with <anchor_vs_one> or <anchor_vs_many> options, respecting to the anchor motif <err*.txt>. For a non-PWM model, the corresponding table should be deduced from the recognition profile of potential hits for the whole genome dataset of promoters of protein-coding genes, e.g. the SiteGA (Tsukanov et al., 2022) tool has a special option to compute the required table

Advanced options include:

• the minimal and maximal spacer lengths restricting the search of spaced locations of motifs The default range of spacer length is [0; 29] nt. MCOT allows the variation of the upper limit of spacer length up to 100 base pairs;
• Expected Recognition Rate (ERR), the frequency of motifs occurrence is required to set the recognition thresholds for motifs, default value 5E-4;
• CE enrichment, threshold –Log10(p-value(CE)) for CE enrichment, default value 5;
• Asymmetry ratio, for two value ERR1 and ERR2 for certain CE means Max{-log10(ERR1), -log10(ERR2)} / Min{-log10(ERR1), -log10(ERR2)}

Motifs recognition

MCOT with options one_partner, many_partners applies the recognition model of PWM for mapping motifs in peaks, otherwise for option anchor_pro MCOT takes ready mapping of predicted hits from a file. For each model, MCOT uses five thresholds {T[1],...T[5]} according to the unified set of ERRs for a whole-genome dataset of promoters, e.g. for the option of command line <pvalue_thr> = 5E-4, five ERRs are equal to {5.24E-5, 1.02E-04, 1.9E-4, 3.33E-4, 5E-4}. The profile of the most stringent hits contains matrix scores T ≥ T[1], the next profile comprises PWM scores {T} in the range T[2] ≥ T > T[1], etc. Hence, MCOT computes five profiles of hits with certain level of conservation for each input motif. Note that the change of the recognition threshold <pvalue_thr> defines the fifth threshold (default value 5E-4) and proportionally shifts the rest four thresholds.

Composite elements search and annotation

MCOT classifies CEs structure according to the following criteria:

• Orientation. Four types of distinct mutual orientations are considered: in the same DNA strand (Direct Anchor/Partner and Direct Partner/Anchor), in opposite strands (Everted and Inverted);
• Overlap or Spacer. There are three distinct cases of mutual locations: Full overlap (one motif located entirely within another one); Partial overlap; and Spacer. To describe each case MCOT uses the following characteristics: the distance between nearest borders of two motifs (Full); the length of overlapped region (Partial); and the length of spacer;
• Asymmetry of сonservation. All predicted CEs are subdivided into two classes: those with more conservative anchor and partner motifs. The conservation of motif hit is estimated as -Log10(ERR), where ERR is computed by the respective score of recognition model

For each of 25 combinations of motifs conservation and each computation flow MCOT compiles the 2x2 contingency table (Table 1) and compute the significance of Fisher’s exact test p-value(CE) that compares the content of CEs in ChIP-seq peaks with that for background model.

Table 1. 2x2 contingency table for calculation of CE significance.

	Sequences without CEs	Sequences with CEs	Total, sequences with both motifs
Observed (peaks)	$Obs_{CE-} = Obs_{Tot} - Obs_{CE+}$	$Obs_{CE+}$	$Obs_{Tot}$
Expected (permuted sequences)	$Exp_{CE-} = Exp_{Tot} - Exp_{CE+}$	$Exp_{CE+}$	$Exp_{Tot}$

The background model implies the preservation of content for each motif in each peak. MCOT generates background profiles of hits of anchor and partner motifs iteratively for each peak with a special permutation procedure.

MCOT computes CE significance separately for five computation flows: Any (Spacer or Overlap), Full, Partial, Overlap (Full and Partial), Spacer. Fisher’s exact test computes the CE enrichment for each of 5x5 combinations of conservation of motifs.

Finally, MCOT estimates the similarity of anchor and partner motifs with the motifs similarity p-value. MCOT used matrix column similarities measures PCC (Pearson Correlation Coefficient, Pietrokovski, 1996) and SSD (Sum of Squared Distances, Sandelin and Wasserman, 2004) to compute two p-values. If maximum among them was less than 0.05, then pair of motifs had the significant similarity. Hence, a respective CE may be a possible false positive prediction.

Additionally, MCOT computed significance of CEs with more conserved anchor motif and more conserved partner motif. These calculations are performed according to scheme represented above in Table 1, but MCOT counts either CEs with more conserved Anchor or Partner motifs, i.e. either -Log10[ERR(Anchor)] > -Log10[ERR(Partner)] or -Log10[ERR(Anchor)] ≤ -Log10[ERR(Partner)].

Finally, to estimate the asymmetry of CE, MCOT partitions all CEs on those with more conserved Anchor or Partner motifs and compute the significance of asymmetry with Fisher’s exact test (Table 2). The asymmetry significance -Log10[P-value] is printed as positive (with sign ‘+’) in the case of enrichment toward the Anchor motif, otherwise, MCOT sign ‘-’ (negative value) denotes the enrichment toward the Partner motif. These calculation yielded for each computation flow one p-value(CE, Asymmetry).

Table 2. 2x2 contingency table for calculation of CE asymmetry.

	CEs with more conserved anchor motif	CEs with more conserved partner motif	CEs, total
Observed (CEs in peaks)	$Obs_{CE, Anchor}$	$Obs_{CE, Partner}$	$Obs_{CE, Anchor} + Obs_{CE, Partner}$
Expected (CEs in permuted sequences)	$Obs_{CE, Anchor}$	$Obs_{CE, Partner}$	$Obs_{CE, Anchor} + Obs_{CE, Partner}$

Detailed enrichment or depletion of CEs with specific combinations of motifs conservation are represented in the scatterplot text file plot_*, see below.

To take into account multiple comparisons we applied the Bonferroni’s correction and used the following critical values to filter out not significant results:

• significance of CEs regardless motifs conservation, Bonferroni_CE = 0.05/(Nfor*Nback*Nflow*Nthr*Nthr);
• significance of asymmetric CEs toward one of motifs, Bonferroni_CE(AncPar) = 0.05/(Nfor*Nback*Nflow*2);
• CE asymmetry, Bonferroni_Asym = 0.05/(Nfor*Nback*Nflow). Here Nfor and Nback means the size of foreground and background datasets (i.e., the number of peaks and random sequences, which generated in MCOT, Nflow = 5 designates the number of MCOT computation flows and Nthr = 5 means the number of thresholds for each motif.

Output data

MCOT gives the following output data:

• Files <*_thr5>, recognition profiles of motifs . Each file respects to one motif. A file has fasta-like format, i.e. for each peak the header line starts with ‘>’ symbol. Next, each subsequent line represents one hit in a peak, particularly it position, respective conservation value -Log10(ERR) and DNA strand.

Example
>Seq 1  Thr 0.864497    Nsites 1
205 3.304404    -
>Seq 2    Thr 0.864497    Nsites 0
>Seq 3    Thr 0.864497    Nsites 2
88    3.607778    -
160    3.338550    -

Here and below ChIP-seq data for mouse FoxA2 and CE FoxA2-HNF1β (Wederell et al., 2008) illustrate MCOT application. The anchor FoxA2 motif we deduced from de novo search Homer (Heinz et al., 2010) and the partner HNF1β motif we took from the mouse Hocomoco core collection (Kulakovskiy et al., 2018).

• Files <err*.txt>, Table Threshold vs. ERR (Expected Recognition Rate) . File contains two columns: threshold and ERR estimated as the site density for the whole genome dataset of protein-coding genes

Example
0.99745670  1.63517e-07
0.99660894  1.88674e-07
0.99612450  4.02504e-07
...
0.81195753  0.0179257
0.81191716  0.0179369
0.81187679  0.0179517

• __File , the detailed recognition statistics__. For each motif and each recognition threshold MCOT provides (1) the number and the name of the motif (anchor motif is designated as ‘Anchor’; numbers 1,2, ... belong to partner motifs), (2) the number and the value of the threshold; (3) the percentage of peaks containing at least one hit of the motif, the number of peaks with recognized motif and the total number of peaks, (4) the number of recognized hits per base pair, the number of recognized hits and the total number of available locations for the motif.

Example

Motif Num	Motif Name	Threshold	% of peaks	Rec. peaks	Total peaks	Rate of hits	Rec. hits	Total positions
0	Anchor	0.955618	32.93	2484	7543	7.23E-04	2887	3994992
0	Anchor	0.945548	19.86	1498	7543	4.20E-04	1677	3994992
0	Anchor	0.934918	25.30	1908	7543	5.50E-04	2199	3994992
0	Anchor	0.923776	29.76	2245	7543	6.69E-04	2672	3994992
0	Anchor	0.913846	28.85	2176	7543	6.63E-04	2647	3994992
1	Partner	0.959810	9.35	705	7543	1.90E-04	760	4002535
1	Partner	0.947124	7.42	560	7543	1.49E-04	596	4002535
1	Partner	0.934158	11.93	900	7543	2.44E-04	978	4002535
1	Partner	0.923150	16.56	1249	7543	3.51E-04	1403	4002535
1	Partner	0.959810	18.60	1403	7543	3.94E-04	1575	4002535

• __File , the summary for statistical significances for all pairs of anchor-partner motifs__ represents the calculation results for different potential CE variants: a homotypic CE (# Motif = Anchor) and one/several heterotypic CE(s) (Partner 1, Partner 2, etc.) for five computation flows (Full/Partial overlap, Overlap, Spacer and Any).

The first block of output data for each pair of motifs represents (a) five P-values of CE enrichment in five computation flows; (b) for each heterotypic pair three P-value for similarity between Anchor and Partner motifs.

Example:

# Motif	Motif Name	Full overlap, -Log10[P-value]	Partial overlap,-Log10[P-value]	Overlap, -Log10[P-value]	Spacer, -Log10[P-value]	Any, -Log10[P-value]	Similarity to Anchor, -Log10[P-value]	Similarity to Anchor, SSD	Similarity to Anchor, PCC	...
Anchor	FOXA2	0.0	97.4	92.4	9.8	35.4	n/a	n/a	n/a
Partner	HNF1B	76.6	53.2	64.3	4.3	35.3	0	0	0

The second block of data consequently for five computation flows represents significances for CEs with more conserved Anchor and Partner motifs.

Example:

# Motif	Motif Name	...	Full overlap, Conservative Anchor, -Log10[P-value]	Full overlap, Conservative Partner, -Log10[P-value]	Partial overlap, Conservative Anchor, -Log10[P-value]	Partial overlap, Conservative Partner, -Log10[P-value]	...
Anchor	FOXA2		n/a	n/a	n/a	n/a
Partner	HNF1B		50.8	0.6	40.1	19.2

The third block of data consequently for five computation flows represents five significances of asymmetry (p-value) in CEs: ‘Anchor vs. Partner’. The positive/negative values reflect the enrichment toward the Anchor/Partner motifs.

Example:

# Motif	Motif Name	...	Full overlap, Asymmetry to Anchor+/Partner-, -Log10[P-value]	Partial overlap, Asymmetry to Anchor+/Partner-, -Log10[P-value]	Overlap, Asymmetry to Anchor+/Partner-, -Log10[P-value]	...
Anchor	FOXA2		n/a	n/a	n/a
Partner	HNF1B		-46.9	-73.1	-113.1

The final fourth block shows recommended Bonferroni’s correction thresholds (-Log10[P-value]) for significances:

# Motif	...	Bonferroni_CE	Bonferroni_CE(AncPar)	Bonferroni_Asym
Anchor	11.8
Partner	11.71	10.61	10.31

• __File <out_hist*>, the abundance of various CE types as a function of mutual orientation and location of the motifs__ The percentage of peaks containing CE variants specific in mutual orientation (four types) and mutual locations from a few possible full overlaps (‘F’), through a variety of partial overlaps (‘P’) and finally from the minimal to the spacer length (‘S’).

Example:

	1F	0F	11P	10P	9P	8P	7P	6P	5P	4P	3P	2P	1P	0S	1S
Everted	2.54	3.88	0.16	0.94	0.07	0.18	0.76	0.09
Inverted	0.07	2.51	0.07	1.08	0.04	0.13	0.72	0.16	0.18	0.31	0.67
DirectPA	0.04	1.86	0.09	0.45	0.02	0.11	0.13	0.09
DirectAP	0.04	0.18	0.07	0.16	0.11	0.02	0.18	0.11	0.11	0.2	0.22
Any	6.84	6.84	0.00	4.77	10.23	3.10	0.29	0.75	3.79	2.24	0.92	1.49	3.51	2.70
Cumulative	6.84	9.60	9.60	14.02	19.43	19.83	20.06	20.34	20.92	21.21	21.95	22.87	23.79	24.66

This example shows distribution for heterotypic CE, since notations DirectAP / DirectPA imply Anchor-Partner / Partner-Anchor cases. A homotypic CE has only one direct orientation of CE, Anchor-Anchor (notation DirectAA). 'Any' row implies frequency of CEs with a certain spacer or an overlap for all four orientations. 'Cumulative' row means the sum of frequencies for all mutual locations which respect to the current and closer positioning of motifs, i.e. the sum of values of 'All' row from the most left column to the current one.

• Files <fisher_*>, the 2x2 tables of CE significance for five computation flows for all motifs pairs and for all 5x5 combination of motifs conservation. Each line of output file contains data concerning one 2x2 contingency table, in particular (1) the designation of conservation (PWM threshold (T), indices from 1 to 5 mean the change from the most stringent to the most permissive, see above); (2) four counts for 2x2 contingency table (see Table 1 above), ‘the number of peaks containing at least one CE (CE+) & ‘the number of peaks containing at least one hit of each motif (Total)’ for peaks (Real) and permuted (Rand) datasets. Finally, the table contains significance of CEs (p-values) computed by Fisher’s exact test for 25 cells of 5x5 tables of combinations of thresholds. Next, the respective data are shown for (a) significances of CEs with more conserved Anchor and Partner motifs (lines 'Anchor', Partner'), significances of any asymmetrical CEs with more conserved either Anchor or Partner motifs (line 'Asymmetry'), significances of symmetrical CEs with the same conservation of Anchor and Partner motifs (line 'Symmetry'), and (b) significances of asymmetry in CEs ‘Anchor vs. Partner’ with the positive/negative Fold respecting to the enrichment toward the Anchor/Partner motifs, significances of asymmetry in CEs ‘Asymmetry vs. Symmetry with the positive/negative Fold respecting to the enrichment toward the Asymmetry/Symmetry in the motifs conservation. In these calculation points (a) and (b) imply counting of peaks and CEs, respectively.

Example below shows FOXA2 (Anchor) and HNF1B (Partner) motifs for Overlap computation flow.

Anchor Thr	Partner Thr		Real CE+	Real Total	Rand CE+	Rand Total	Fold	P-value
A 1	P 1		13	211	20	14137	43.550	4.59308e-16
A 1	P 2		10	167	12	11189	55.833	2.00465e-13
A 1	P 3		9	142	18	9514	33.500	9.33201e-11
A 1	P 4		22	312	27	20904	54.593	7.98321e-28
A 1	P 5		32	359	38	24053	56.421	2.97654e-40
A 2	P 1		10	125	18	8375	37.222	3.45082e-12
A 2	P 2		13	78	8	5226	108.875	9.83765e-20
A 2	P 3		10	91	5	6097	134.000	8.07117e-16
A 2	P 4		21	172	23	11524	61.174	1.41932e-27
A 2	P 5		23	191	25	12797	61.640	4.1353e-30
A 3	P 1		7	160	18	10720	26.056	5.01905e-08
A 3	P 2		11	106	17	7102	43.353	7.15405e-14
A 3	P 3		15	107	18	7169	55.833	9.86635e-20
A 3	P 4		18	229	37	15343	32.595	4.69822e-20
A 3	P 5		23	263	32	17621	48.156	3.30391e-28
A 4	P 1		11	192	35	12864	21.057	4.44823e-11
A 4	P 2		11	133	13	8911	56.692	9.7153e-15
A 4	P 3		24	124	29	8308	55.448	5.57973e-31
A 4	P 4		36	303	58	20301	41.586	7.52749e-42
A 4	P 5		37	286	39	19162	63.564	5.87211e-48
A 5	P 1		17	171	13	11457	87.615	3.18051e-24
A 5	P 2		20	124	15	8308	89.333	1.23212e-28
A 5	P 3		80	151	59	10117	90.847	1.07118e-118
A 5	P 4		55	229	37	15343	99.595	7.19329e-79
A 5	P 5		53	250	33	16750	107.606	9.56812e-77
Anchor			170	1392	237	93264	48.059	1.76431e-199
Partner			95	1057	106	70819	60.047	3.61688e-118
Asymmetry			399	1741	598	116647	44.704	1e-300
Symmetry			0	0	0	0	1.000	1
Anchor_Partner			280	1418	335	904	0.533	7.08806e-20
Asym_Sym			1418	1418	904	904	1.000	2

• Files <*.best>, the list of predicted CEs. For each recognized CE MCOT provides (1) the header of a peak, (2) the start and the end positions of each motif in a peak, (3) mutual location (Full / Partial / Spacer types and the respective base pair measures, see above), (4) the strands of the Anchor/Partner motifs in a peak and the mutual orientation of the motifs (one of four types), (5) conservation scores and DNA sequences of the motifs. To provide detailed information on asymmetry ‘Anchor vs. Partner’ in comparison of peaks and permuted sequences MCOT provides files <real.best> and rand.best> that show the lists of predicted CE for peaks and permuted sequences.

#Seq	Anchor start	Anchor end	Partner start	Partner end	Mutual Loc	Loc Type	Strands	Mutual Ori	Anchor hit conservation, -Log10(ERR)	Partner hit conservation, -Log10(ERR)	Anchor seq	Partner seq
Seq 18	271	282	268	282	0F	Full	-+	Inverted	3.365	3.447	tgtttatctttc	agtgaaagataaaca
Seq 23	565	576	569	583	8P	Partial	-+	Everted	3.457	4.506	tattgacttacc	agtcaataagttaca
Seq 33	145	156	179	193	22S	Spacer	++	DirectAP	3.448	4.289	tgttgacagact	ggttaatgctttcct

• Files <plot_*>, heatmaps that show the CE asymmetry, i.e. the abundance of CEs with various ratios of conservation of Anchor and Partner motifs For each of five computation flows (Full, Partial, Overlap, Spacer and Any) one heatmap is computed. For foreground and background data (peaks and sequences with permuted hits) the respective list of predicted CEs are subdivided on two fractions: those with more conserved Anchor and Partner motifs. The conservation of a hit is estimated with the respective -Log10(ERR) value. The minimal conservation value is equal to -Log10(5E-4) ~ 3.3. Next, MCOT computes two matrices {OBSi,j} and {EXPi,j} of absolute frequencies of conservation of Anchor and Partner motifs for observed and expected data. Here indices i and j imply the conservation -Log10(ERR) of Anchor and Partner motifs. E.g., for ERR = 5E-4 (-Log10(5E-4) ~ 3.30) this conservation is falling within the ranges [<3.5], [3.5..3.7], [3.7..3.9] etc. up to [5.3..5.5] and [>5.5]. Finally, the per mille measure transforms the absolute frequencies to relative ones as follow:

${1000*OBS_{i,j}/N(OBS)}$ and ${1000*EXP_{i,j}/N(EXP)}$,

where $N(OBS)$ and $N(EXP)$ are total counts of predicted CEs in observed and expected lists. The output heatmap shows the difference between relative frequencies of observed and expected CEs. Example below shows the asymmetry toward the Partner (HNF1B) motif in Overlap computation flow.

	<3.5	3.5..3.7	3.7..3.9	3.9..4.1	4.1..4.3	4.3..4.5	4.5..4.7	4.7..4.9	4.9..5.1	5.1..5.3	5.3..5.5	>5.5
<3.5	14	14	-8	-14	-26	-12	1
3.5..3.7	7	-15	-7	-7	-2
3.7..3.9	37	-4	-2	-6	4	-11	-3	-3
3.9..4.1	17	-1	14	1	-4	-9
4.1..4.3	39	-2	8	5	-8	-2
4.3..4.5	3	-6	4	5	5
4.5..4.7	-10	22	-5
4.7..4.9	13	-4	-3
4.9..5.1	-23	-3	5	-4	1
5.1..5.3	1	-2
5.3..5.5	-4	-3	-2	-3
>5.5	4	2	-1	-2	-3

The same calculations are performed for Anchor-Anchor CEs, in this case the enrichment of symmetrical vs. asymmetrical CEs are tested.

References

Bailey, T.L. (2021) STREME: accurate and versatile sequence motif discovery. Bioinformatics. 37, 2834–40

Levitsky, V., Zemlyanskaya, E., Oshchepkov, D., Podkolodnaya, O., Ignatieva, E., Grosse, I., et al. (2019) A single ChIP-seq dataset is sufficient for comprehensive analysis of motifs co-occurrence with MCOT package. Nucleic Acids Res. 47, e139.

Levitsky, V., Oshchepkov, D., Zemlyanskaya, E., Merkulova, T. (2020) Asymmetric conservation within pairs of co-occurred motifs mediates weak direct transcription factor binding in ChIP-seq data. Int J Mol Sci. 21, 6023.

Levitsky, V.G., Mukhin, A.M., Oshchepkov, D.Y., Zemlyanskaya, E.V., Lashin, S.A. (2022) Web-MCOT Server for Motif Co-Occurrence Search in ChIP-Seq Data. Int. J. Mol. Sci., 23, 8981.

Heinz, S., Benner, C., Spann, N., Bertolino, E., Lin, Y.C., Laslo, P., et al. (2010) Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol Cell, 38, 576-589.

Kulakovskiy, I.V., Vorontsov, I.E., Yevshin, I.S., Sharipov, R.N., Fedorova, A.D., Rumynskiy, E.I., et al. (2018) HOCOMOCO: expansion and enhancement of the collection of transcription factor binding sites models. Nucleic Acids Res., 46, D252-D259.

O'Malley, R.C., Huang, S.C., Song, L., Lewsey, M.G., Bartlett, A., Nery, J.R., et al. (2016) Cistrome and epicistrome features shape the regulatory DNA landscape. Cell, 165, 1280-1292.

Pietrokovski, S. (1996) Searching databases of conserved sequence regions by aligning protein multiple-alignments. Nucleic Acids Res., 24, 3836-3845.

Sandelin, A., Wasserman, W.W. (2004) Constrained binding site diversity within families of transcription factors enhances pattern discovery bioinformatics. J Mol Biol., 338, 207-215.

Siebert, M., Söding, J. (2016). Bayesian Markov models consistently outperform PWMs at predicting motifs in nucleotide sequences. Nucleic Acids Res., 44, 6055–69.

Tsukanov, A. V., Mironova, V. V., Levitsky, V. G. (2022) Motif models proposing independent and interdependent impacts of nucleotides are related to high and low affinity transcription factor binding sites in Arabidopsis. Front Plant Sci., 13, 938545.

Wederell, E.D., Bilenky, M., Cullum, R., Thiessen, N., Dagpinar, M., Delaney, A., et al. (2008) Global analysis of in vivo Foxa2-binding sites in mouse adult liver using massively parallel sequencing. Nucleic Acids Res., 36, 4549-4564.