Experimental procedures

Construction of the TSS-seq Libraries and Analysis of the TSS Tags


	Two hundred micrograms of the obtained total RNA were subjected to oligo-capping with some modifications from the original protocol; namely, after the successive treatment of the RNA with 2.5 U BAP (TaKaRa) at 37 °C for 1 hour and 40 U TAP (Ambion) at 37 °C for 1 hour, the BAP-TAP-treated RNAs were ligated with 1.2 micrograms of the following RNA oligo (5'- AAUGAUACGGCGACCACCGAGAUCUACACUCUUUCCCUACACGACGCUCUUCCGAUCUGG-3'), using 250 U T4 RNA ligase (TaKaRa) at 20 °C for 3 hours. After the DNase I treatment (TaKaRa), polyA-containing RNA was selected using oligo-dT powder ,(Collaborative). The first strand cDNA was synthesized from 10 picomoles of random hexamer primer (5'-CAAGCAGAAGACGGCATACGANNNNNNC-3') using Super Script II (Invitrogen), by an incubation at 12 °C for 1 hour and at 42 °C overnight. The template RNA was degraded by an alkaline treatment. For PCR, a 20% portion of the first strand cDNAs was used as the PCR template. Gene Amp PCR kits (PerkinElmer) were used with the PCR primers 5'-AATGATACGGCGACCACCGAG-3' and 5'-CAAGCAGAAGACGGCATACGA-3',under the following reaction conditions: 15 cycles of 94 °C for 1 minute, 56 °C for 1 minute and 72 °C for 2 minutes,The PCR fragments were size-fractionated by 12 % polyacrylamide gel electrophoresis, and the fraction containing the 150-250 bp fragments was recovered.The quality and quantity of the obtained single-stranded first strand cDNAs were assessed, using a BioAnalyzer (Agilent). One nanogram of the size-fractionated cDNA was used for the sequencing reactions with the Illumina GA. The sequencing reactions were performed according to the manufacturer's instructions. The 36-bp read TSS tags were mapped onto the human genome sequence (hg18, UCSC Genome Browser) using ELAND. Uniquely and completely (without any mismatches) mapped TSS tags were used for the analysis. After the clustering of the mapped TSS tags into the 500-bp bin, the TSS clusters located within the region from -50 kb of the 5'-end to the 3'-end boundary of the RefSeq gene were selected. TSS clusters were removed when all of the belonging TSS tags were located at the internal exonic region of the corresponding RefSeq genes. The RefSeq information, such as the genomic coordinates, the position of the protein coding region, etc, is from hg18. Gene Ontology terms were correlated with RefSeq, using loc2go. Briefly, TSS-seq combines our full-length cDNA technology, oligo-capping (Suzuki and Sugano, Methods in Mol Biol, 2003), and Illumina GA technology. Adaptors necessary for Illumina sequencing is introduced to the cap site by three steps of enzymatic reactions, which enables to determine the immediately downstream sequences of the TSSs (TSS tag). Adaptors containing necessary sequence for the Illumina GA sequencer are represented as gray boxes. For further information, see the reference. Gppp: cap structure. AAA: polyA

QC Information

Experimental procedures

IHEC data

House kepping genes informations
Download IHEC exon intron map Information
Download IHEC Summary House keeping genes expressions
Download IHEC All House keeping genes expressions

Statistics of IHEC RNA-seq.

Sample ID	exon mapped		intron mapped


COC4	72,402,402	68%	6,767,491	6%	26,311,725	24%
COC5	76,007,419	71%	6,914,205	6%	23,978,825	22%
COC6	43,012,847	71%	5,736,689	9%	11,157,621	18%
COC7	68,562,432	67%	9,872,805	9%	22,539,376	22%
COC8	42,924,747	61%	8,158,582	11%	18,571,310	26%
COC11	62,607,978	65%	5,658,141	5%	27,585,868	28%
COC14	65,661,920	71%	6,338,731	6%	19,795,450	21%
COC15	44,593,728	67%	6,987,910	10%	14,940,597	22%
COC16	47,820,333	70%	7,784,506	11%	12,228,659	18%
COC20	53,256,053	68%	6,493,434	8%	17,598,691	22%
COC21	67,117,265	70%	4,831,113	5%	23,303,757	24%
COC22	73,183,585	75%	5,009,258	5%	18,938,413	19%
HPC6	46,270,832	67%	5,175,776	7%	17,111,849	24%
HPC8	96,406,613	67%	12,392,091	8%	33,806,117	23%
HPC17	96,459,793	73%	8,293,334	6%	27,147,997	20%
HPC20	103,557,369	76%	11,647,843	8%	20,970,591	15%
HPC25	95,494,229	62%	14,359,252	9%	43,086,227	28%
HPC27	95,723,872	69%	11,616,068	8%	31,341,884	22%
HPC28	115,334,113	73%	12,689,342	8%	28,725,557	18%
HPC35	81,811,570	75%	7,144,859	6%	19,586,719	18%
JKU001	291,008,424	84%	25,095,294	7%	27,878,805	8%
JKU002	260,157,542	83%	22,973,483	7%	27,700,571	8%
JKU003	225,385,508	82%	19,387,155	7%	28,427,719	10%
JKU004	222,561,799	86%	12,867,676	4%	22,257,456	8%
JKU005	196,717,060	70%	26,924,308	9%	54,654,832	19%
JKU006	217,777,797	79%	26,241,531	9%	31,517,599	11%
JKU007	184,687,741	85%	10,474,785	4%	21,704,775	10%
JKU008	188,386,330	85%	11,958,713	5%	19,565,296	8%
JKU009	269,444,863	88%	11,591,228	3%	22,607,829	7%
JKU011	242,044,367	92%	6,567,446	2%	12,159,464	4%
JKU013	210,444,383	91%	6,472,297	2%	12,948,991	5%
JKU014	209,603,080	92%	5,192,995	2%	11,117,491	4%
JKU015	236,846,854	86%	13,234,256	4%	24,767,411	9%
JKU016	76,782,724	86%	7,095,077	7%	5,376,020	6%
JKU017	76,618,309	86%	6,100,492	6%	5,845,636	6%
JTK001	43,051,071	79%	5,189,228	9%	6,180,297	11%
JTK002	44,694,507	82%	3,635,373	6%	5,768,031	10%
JTK003	49,924,561	82%	4,616,985	7%	6,038,998	9%
JTK004	46,775,574	82%	4,014,663	7%	5,857,107	10%
JTK005	101,254,531	79%	11,087,306	8%	14,999,865	11%
JTK006	130,295,760	79%	14,164,041	8%	20,426,786	12%
JTK007	45,875,725	79%	4,921,764	8%	6,685,829	11%
JTK008	42,841,666	82%	3,782,099	7%	5,274,874	10%
JTK009	39,114,123	82%	3,301,281	6%	4,928,929	10%
JTK010	38,561,358	82%	3,239,226	6%	4,869,975	10%
JTK011	41,312,008	82%	3,415,169	6%	5,153,142	10%
JTK012	45,216,074	83%	3,385,427	6%	5,330,681	9%
JTK013	105,622,271	78%	11,593,014	8%	17,173,491	12%
JTK014	52,512,954	81%	4,533,195	7%	7,488,527	11%
JTK015	222,750,215	82%	17,008,792	6%	29,287,851	10%
JTK016	61,588,390	81%	5,275,061	6%	9,071,050	11%
JTK017	29,570,988	81%	2,714,317	7%	4,001,747	11%
JTK018	59,367,525	76%	8,506,824	10%	9,576,918	12%
JTK019	52,814,137	76%	7,702,326	11%	8,535,127	12%
JTK020	108,554,869	74%	18,187,001	12%	18,369,691	12%
JTK021	81,278,938	77%	10,849,982	10%	12,571,906	12%
JTK022	43,117,644	77%	5,225,899	9%	7,197,539	12%
JTK023	34,932,088	78%	3,956,775	8%	5,488,247	12%
JTK024	83,780,897	76%	11,591,656	10%	14,856,879	13%
JTK025	68,724,536	76%	9,460,328	10%	11,369,100	12%
JTK026	105,117,353	80%	9,534,630	7%	16,548,971	12%

Drug perturbation of lung cancer

Statistics of Drug perturbation RNA-seq.

		Total data points∗	Average numbers per data point

		Total data points∗	Total reads	Mapped reads	%mapped	%intron in mapped

Dataset-1
	A549	249	2,129,706	1,455,347	68%	8%
	H1299	253	1,823,507	1,275,695	69%	7%
	H1648	251	1,934,265	1,319,718	68%	7%
	H2347	245	1,998,116	1,414,144	70%	9%
	II-18	231	1,985,839	1,433,442	72%	10%

Dataset-2	23cells	2011	1,794,226	1,295,897	72%	9%

Statistics of Drug perturbation ATAC-seq.

		Total data points∗	Average numbers per data point

		Total data points∗	Mapped reads	%mapped	ChrM rmapped	%chrM in mapped	MACS peaks

Dataset-1
	A549	251	11,286,126	79%	2,287,959	17%	35,355
	H1299	269	5,821,704	69%	3,861,822	65%	17,184
	H1648	264	5,943,835	71%	3,510,469	58%	18,571
	H2347	276	4,752,301	70%	3,089,295	64%	26,166
	II-18	256	5,838,263	70%	3,849,410	64%	18641

Dataset-2	23cells	2077	8,158,509	71%	5,304,584	57%	17,734

∗>0.5 million total reads; spike-in control within 2sd;<15 ‰ intron reads

Basal multi-omics data in 26 cell lines

Download QC Informatino

House kepping genes informations
Download Base multi omics 26 Cell RNA-seq exon intron map Information
Download Base multi omics 26 Cell RNA-seq Summary House keeping genes expressions
Download Base multi omics 26 Cell RNA-seq All House keeping genes expressions

・Whole Genome-seq

	Mapped sequences (Read1+Read2)	Depth (avg)	Coverage (x5)


PC-7	1,181,752,959	38.4	0.92
PC-9	1,235,410,075	40.2	0.91
PC-14	1,377,953,696	44.8	0.91
RERF-LC-ad1	1,189,406,566	38.7	0.91
RERF-LC-ad2	1,204,958,194	39.2	0.92
RERF-LC-KJ	1,058,371,222	34.4	0.92
RERF-LC-MS	1,234,238,416	40.2	0.92
RERF-LC-OK	677,038,144	21.8	0.91
VMRC-LCD	1,270,060,339	41.3	0.91
ABC-1	1,131,071,585	36.8	0.91
LC2/ad	1,265,338,449	41.2	0.91
II-18	860,643,037	27.6	0.90

A549	723,563,287	22.2	0.86
A427	1,040,002,036	33.8	0.91
H322	893,332,828	28.9	0.90
H2228	830,781,519	27.0	0.91
H1299	899,909,551	29.3	0.91
H1437	711,909,693	23.1	0.91
H1648	1,065,042,096	34.6	0.92
H1650	1,031,008,238	33.5	0.91
H1703	984,465,974	31.6	0.91
H1819	1,091,791,039	35.5	0.91
H1975	1,004,161,315	32.6	0.91
H2126	640,653,382	20.8	0.91
H2347	948,973,026	30.9	0.91

・RNA-seq

Cell line	Used sequences (Read1)	Num of genes


		> 1 RPKM	> 5 RPKM

PC-3	49,914,547	12,205	9,240
PC-7	50,925,975	12,129	9,009
PC-9	34,167,521	12,817	9,532
PC-14	53,977,381	12,169	9,037
RERF-LC-Ad1	56,406,046	12,298	9,206
RERF-LC-Ad2	45,580,359	12,392	8,804
RERF-LC-KJ	60,803,665	12,054	8,938
RERF-LC-MS	52,715,099	13,045	9,090
RERF-LC-OK	33,086,988	12,309	8,954
VMRC-LCD	45,944,953	12,502	8,711
ABC-1	37,993,504	11,715	8,384
LC2/ad	43,665,988	12,366	9,206
II-18	63,869,445	11,955	9,038

A549	20,440,396	12,155	8,998
A427	41,895,881	11,866	9,011
H322	54,487,583	12,457	9,351
H2228	56,465,940	12,409	9,106
H1299	51,120,991	11,735	8,958
H1437	49,890,034	12,275	8,921
H1648	38,908,100	12,604	9,317
H1650	26,635,691	12,716	9,595
H1703	87,705,180	11,736	8,695
H1819	75,262,673	12,494	9,185
H1975	36,195,247	12,715	9,634
H2126	46,862,796	12,143	9,016
H2347	50,325,156	12,278	9,030
SAEC	180,054,144	12,126	8,809

・qRT-PCR → Supplementary Fig. S2(Suzuki et al. 2014 NAR)
・RT-PCR for fusion transcripts → Supplementary Fig. S13 (Suzuki et al. 2014 NAR)
For expression abundances, we conducted 1,352 qRT-PCR assays (52 genes × 26 cell lines; N=3).
The rpkm values of RNA-seq were positively correlated with the Ct values of qRT-PCR (R= 0.89).
We also validated several fusion transcripts detected in this study using RT-PCR.

・ChIP-seq
ChIP-seq 26 lung cancer cell lines

	Average of mapped sequences	Average of number of peaks (MACS2)


		Narrow peaks	Broad peaks

H3K4me3	26,140,455	21,209	16,208
H3K9/14ac	19,596,187	34,374	23,753
Pol II	26,056,772	15,715	13,997
H3K36me3	24,264,604	107,708	47,710
H3K4me1	25,900,257	108,882	75,854
H3K27ac	25,690,276	61,061	38,297
H3K27me3	21,584,812	53,587	42,163
H3K9me3	21,155,573	39,559	51,760
WCE	19,100,553	∗∗∗	∗∗∗

ChIP-seq SAEC

	Average of mapped sequences	Average of number of peaks (MACS2)


		Narrow peaks	Broad peaks

H3K4me3	43,579,277	15,626	14,093
H3K9/14ac	21,603,337	50,674	45,159
Pol II	21,986,637	16,703	15,234
H3K36me3	56,493,935	321,485	107,145
H3K4me1	52,851,492	226,330	154,297
H3K27ac	45,848,952	170,013	88,659
H3K27me3	29,626,299	88,943	83,095
H3K9me3	40,496,823	316,142	148,544
WCE	45,429,763	***	***

・qPCR (pre-sequencing QC) → Supplementary Table S4 (Suzuki et al. 2014 NAR)
・qPCR of cancer-related genes → Supplementary Fig. S4 (Suzuki et al. 2014 NAR)
・Replicate → Supplementary Fig. S5 (Suzuki et al. 2014 NAR)
・Comparison with ENCODE A549 H3K4me3 → Supplementary Fig. S6 (Suzuki et al. 2014 NAR)

Before sequencing ChIP samples, we conducted qPCR quality control by analyzing ChIP signals in positive control and negative control regions.
We also performed qPCR validations of 65 assays (N=3) for cancer-related genes.
The intensities of ChIP-seq data were moderately correlated with those of qPCR (R=0.63).
The results of H3K4me3 and H3K27ac ChIP-seq (peak type) showed strong positive correlations with qPCR results (R = 0.84 and 0.74).
The correlation of repressive marks (H3K27me3 and H3K9me3; broad type) were just moderately correlated (R = 0.46).
We also performed ChIP experiments twice to confirm reproducibility of ChIP-seq profiles.
The signal intensities showed strong correlation (R = 0.998, H3K4me3 of H1975; R = 0.946, Pol II of LC2/ad).
We also compared our ChIP-seq data (A549 H3K4me3) with ENCODE data (wgEncodeEH001905 and wgEncodeEH001904).About 88% of the peaks were overlapped and the signal intensities in promoters were strongly correlated between our dataset and the ENCODE datasets (R=0.96).

BS-seq

Cell line	Mapped sequences	Avg. of depths	Conversion rate (x5)^∗	CpG sites (> x5)


PC-3	157,902,653	161.4	0.994	3,673,159
PC-7	109,919,011	110.9	0.994	3,418,929
PC-9	87,012,056	89.6	0.994	3,231,320
PC-14	204,216,479	210.3	0.994	4,064,068
RERF-LC-Ad1	87,043,746	89.1	0.992	3,264,395
RERF-LC-Ad2	78,300,691	83.0	0.994	3,448,211
RERF-LC-KJ	72,844,738	74.9	0.993	3,068,971
RERF-LC-MS	102,938,936	109.0	0.994	3,598,662
RERF-LC-OK	161,552,507	165.0	0.993	3,758,532
VMRC-LCD	84,681,570	89.5	0.992	3,136,774
LC2/ad	112,097,386	116.0	0.988	3,548,548
ABC-1	93,158,547	93.1	0.993	3,493,903
II-18	99,682,438	165.0	0.993	3,327,001

A549	87,966,180	91.0	0.991	3,324,364
A427	53,499,542	54.3	0.992	2,614,641
H322	153,896,186	165.8	0.989	4,161,775
H2228	122,705,759	81.6	0.993	4,815,543
H1299	118,923,875	82.2	0.994	4,533,930
H1437	98,311,209	63.1	0.993	4,382,225
H1648	102,033,841	104.4	0.989	3,357,747
H1650	105,694,196	109.4	0.994	3,460,378
H1703	127,897,486	81.6	0.994	5,513,896
H1819	220,008,485	223.4	0.986	4,085,231
H1975	79,688,628	81.7	0.993	3,274,116
H2126	124,651,437	80.2	0.993	4,991,289
H2347	115,973,241	76.1	0.993	4,661,415

・Conversion rate → Supplementary Table S11 (Suzuki et al. 2014 NAR)
・Sanger sequencing validation (direct sequencing and TA cloning) → Supplementary Fig. S3 and Table S3 (Suzuki et al. 2014 NAR)
For quality control of bisulfite conversion, C to T conversion rates were calculated at C of non-CpG sites.
All of the BS-seq data satisfied more than 99% of the conversion rate.
To validate DNA methylation patterns of CpG sites, direct Sanger sequencing of BS-seq libraries were performed for several regions.
We also conducted TA cloning of 12 assays for seven genes. As a result, 83% of the CpG sites were precisely validated by Sanger sequencing validations.

Single-cell sequencing C1 (Fluidigm)
・Spike-in control → Fig. 1A (Suzuki et al. 2015 Genome Biology)
・qPCR → Fig. 1C, Fig. 2E, Fig. 2F (Suzuki et al. 2015 Genome Biology)
・Replicate → Fig. 1D (Suzuki et al. 2015 Genome Biology)
All single-cell libraries of C1 included three RNA spike-in controls. we removed the libraries in which tag counts of any of the spike-in controls deviated by more than 2 sd. We conducted qPCR validations using cDNA samples of individual single cells. The results of single-cell qPCR assays were highly correlated with those of bulk (200 cells) qPCR assays (R=0.94). Furthermore, we compared expression levels and these relative divergences of single-cell RNA-seq (scRNA-seq) and qPCR results and confirmed strong correlation between them (R=0.87 and R=0.84, respectively). For reproducibility of the scRNA-seq data, we repeated scRNA-seq experiments twice using LC2/ad (R=0.93). The average expression levels of the sequencing replicates were strongly correlated (R=0.99). We also compared average expression levels of scRNA-seq with those of bulk datasets (R=0.86, 200 cells with the same protocol with scRNA-seq; R=0.82, 107 cells with different protocol of the library preparation). All of the C1 data and figures for the QC and validation study were shown in Suzuki et al. Single-cell analysis of lung adenocarcinoma cell lines reveals diverse expression patterns of individual cells invoked by a molecular target drug treatment 2015 Genome Biology.

Workflow of data processing for the 26 lung adenocarcinoma cell lines

Experimental procedures

Whole-genome sequencing

The paired-end sequences were mapped to the human reference genome using the Burrows-Wheeler Aligner (BWA, v0.6.0-r85). PCR duplicates were discarded using SAMtools version 0.1.18 (r982:295). SNVs and indels were identified using the Genome Analysis Toolkit (GATK, v1.6-5-g557da77) Unified Genotyper or Somatic Indel Detecter, respectively. After filtering SNVs and indels, germline mutations registered in the public databases and in-house catalogues of Japanese normal variations were removed, and somatic mutations in COSMIC v59 were rescued.
*Sufficient supporting tags: for SNVs, variant tags > 4; for indels, variant tags > 4, variant tags (Fwd) > 1 and variant tags (Rev) > 1.
**NCBI dbSNP build 137, Exome Sequencing Project (ESP6500SI-V2) (AF > 0.1%), the 1000 Genomes Project (phase1_v3, downloaded on 2013.10.10) (AF > 0.1%) and in-house Japanese SNP data from 145 Japanese normal tissues.
CNA (Copy Number Aberration; ploidity) was called using Control-FREEC (v3.4) with the parameters of ploidy=2, window=1,500.

RNA sequencing

For calculation of expression abundancies, RNA-seq data was mapped to the reference genome using ELAND (illumina). The ppm and rpkm values were calculated using our in-house manuscripts.
To detect fusion transcripts, sequences were mapped using TopHat2 (v2.0.6) and fusion candidates were detected by TopHat-fusion and tophat-fusion-post.
Finally, 135 fusion transcript candidates were identified.
*TopHat2 options: --bowtie1, --mate-std-dev 80, --max-intron-length 100000, --fusion-min-dist 10000000 and --fusion-anchor-length 13.
** Sufficient qualities for fusion transcripts: number of spanning reads >10; number of spanning mate pairs >2 and number of spanning mate pairs where one end spans a fusion >2.
For visualization and detection of exon-intron junctions, paired-end sequences were mapped using TopHat2 (v2.0.6) with the option -r 50.

ChIP sequencing

ChIP-seq data was mapped to the reference genome using ELAND. For peak calling, we used MACS2 with default parameters.
We also calculated signal intensities of promoter and enhancer regions by counting ChIP-seq tags and normalizing the signals of ChIP samples with those of whole cell extract (WCE) control samples.

To define chromatin patterns combining all ChIP-seq data, we conducted ChromHMM analysis and defined eight chromatin state, namely active promoter, weak/poised promoter, strong enhancer, weak enhancer, transcriptional elongation, inactive region, inactive region/heterochromatin, and low/no signal.

Bisulfite sequencing

We modified bisulfite sequencing (BS-seq) data as below; read1: C to T, read2: G to A. Using the BWA (v0.6.0-r85), the modified sequences were mapped to the modified reference genome (G to A).
According to these mapping result, pre-modified sequences were mapped.
The number of CG, CA, CT and CC with methylated C and TG, TA, TT and TC with non-methylated C were counted.
Ratios of methylated C to the depth in each CpG site were calculated as DNA methylation rates.

Single-cell sequencing


Sufficient qualities for C1 libraries: >2 million mapped reads and spike-in control within 2 sd. *Sufficient qualities for Chromium libraries: >5,000 tags .

Long-read sequencing