¹⁵NH₄Cl was purchased from Cambridge Isotope Laboratories, Andover, MA, United States. M. smegmatis MC² 155 was cultured in an M9 minimal medium with ¹⁵N metabolite labeling described previously (Zhu et al., 2022). Briefly, the ¹⁴NH₄Cl- and ¹⁵NH₄Cl -labeled cells were equally mixed and disrupted with a Soniprep sonicator (Scientz, Ningbo, China) for 15 min (2 s-on, 4 s-off) at 30% amplitude. Supernatants were collected after centrifugation at 13,000 rpm for 15 min, and protein concentration was measured by a gel-assisted method as described previously (Zhang et al., 2016b).

To achieve deep coverage of M. smegmatis proteome, we reduced 240 μg of proteins with 5 mM of dithiothreitol (DTT) and alkylated them with 20 mM of iodoacetamide (IAA). The alkylated proteins were split into two samples, separated by a 10% SDS-PAGE for 8 cm, and stained with Coomassie Blue G250. Two gel lanes were excised into 13 fractions based on the molecular weight (MW) and protein abundance and digested with Ac-Trypsin (Wu et al., 2016; 12.5 ng/μL) and Ac-LysargiNase (Zhang et al., 2019; 12.5 ng/μL) at 37°C for 12–24 h. The extracted peptides were desalted with a homemade C₁₈ StageTip (Zhai et al., 2013), dried, and dissolved in loading buffer (1% acetonitrile, ACN, and 1% formic acid, FA) for MS analysis.

The dissolved peptides (500 ng) were analyzed, as described previously. Briefly, the liquid chromatography-tandem mass spectrometry (LC-MS/MS) consisted of an EASY-nLC 1200 system (Thermo Fisher Scientific, San Jose, CA, United States) equipped with a self-packed capillary column (75 μm i.d. × 15 cm, 3 μm C₁₈ reversed-phase fused silica) coupled to an Orbitrap Fusion Lumos (Thermo Fisher Scientific). For full MS scans, the automatic gain control (AGC) was set at 5.0 × 10⁵. The scan ranged from 300 to 1,400 m/z at a resolution of 1.2 × 10⁵ and a maximum injection time (MIT) of 50 ms. For the MS₂ scan, only spectra with a charge state of 2–6 were selected for fragmentation by higher energy collision-induced dissociation (HCD) with a normalized collision energy of 32%, an AGC of 1 × 10⁵, and an MIT of 35 ms. The dynamic exclusion was set to 30 s.

The annotated protein database of M. smegmatis MC² 155 was downloaded from NCBI (NZ_CP009494, https://www.ncbi.nlm.nih.gov/genome/1026?genome_assembly_id=212020), including 6,385 entries. Using pAnno (Zhang et al., 2015) in pFind software, the six-frame database was constructed based on the complete genome sequence of MC² 155 according to a stop-to-stop translating strategy [183,120 entries with at least seven amino acids (AAs)]. Two MS/MS datasets digested with different proteases were analyzed using pFind (v 3.1.5; Chi et al., 2018) to search against the annotated and six-frame databases. For the two digestion modes, trypsin and lysargiNase, a maximum of two missed cleavages was tolerated; 20 ppm was tolerated in MS and MS/MS search modes. Cysteine carbamidomethyl was set as a fixed modification, whereas acetylation of protein N-termini and methionine oxidation were added as variable modifications. A minimal peptide length of 7 AAs was required. The false discovery rates (FDR) for the peptide-spectrum match (PSM), peptide, and protein were set to < 1%. ¹⁵N labeling mode was chosen for novel peptide verification.

The identified peptides were from three types: annotated proteins, novel ORFs, and N-terminal corrections according to the encoding gene position. The potential novel peptides mapping to the N-terminal extension of the annotated proteins and the unannotated regions were obtained and evaluated for their accuracy by “mirror proteases” digesting the evidence, Q-value, raw score, and ¹⁴N and ¹⁵N spectrum similarity. A Q-value was calculated based on the target-decoy approach. A raw score is the reliability of the PSM. ¹⁴N and ¹⁵N spectrum similarity was calculated by pBuild in pFind software, as described previously (Zhu et al., 2022). In addition, novel peptides with higher, moderate, and lower scores were selected and synthesized for further verification. The spectrum similarity of original and synthesized peptides was checked. A Student's t-test was calculated by R (^***P < 0.001, ^**P < 0.01, ^*P < 0.05).

The public RNA-seq (SRR17866681, 2-Feb-2022) and Ribo-seq (E-MTAB-2929, 1-Jan-2015; Shell et al., 2015) datasets were downloaded from NCBI and ArrayExpress, respectively, and aligned to reference DNA sequence by HISAT2 (v 2.2.1; Kim et al., 2019), and SAMtools (v 1.11; Li et al., 2009). The expression of RNA-seq and Ribo-seq was calculated by Cufflinks (v2.2.1; Trapnell et al., 2012) locally and visualized by IGV (v 2.11.1; Thorvaldsdottir et al., 2013) and R (v 4.1.2).

A total of 150 μg proteins were reduced by 10 mM dithiothreitol (DTT) at 45°C for 60 min and alkylated by 20 mM IAA for 45 min at room temperature. Then, 40 mM formaldehyde (Sigma-Aldrich, Saint Louis, MO, United States) and 20 mM sodium cyanoborohydride (Sigma-Aldrich) were added to the above protein sample and incubated at 37°C overnight. A solution of 1 M Tris (pH 6.8) with a final concentration of 0.1 M was added to quench the reaction.

The dimethylated proteins were digested with Ac-Trypsin at a protein/enzyme ratio of 50:1 (w/w) at 37°C for 14 h. The high molecular weight dendritic hyperbranched polyglycerol-aldehydes (HPG-ALD, Vancouver, BC, Canada; Kleifeld et al., 2010) polymer was added to the peptide sample. 20 mM sodium cyanoborohydride was immediately added and incubated overnight at 37°C. The naturally blocked and experimentally labeled N-terminal peptides were collected by ultrafiltration with the 30 kDa MWCO Amicon column (Sartorius, Gottingen, Germany).

The N-terminal peptide sample was fractionated by the RP-Tip (Reverse-Phase Tip; Ni et al., 2019) with an increasing acetonitrile step-gradient (6, 9, 12, 15, 18, 21, 25, 30, 35, and 50%, pH = 10). These 10 fractions were combined into six samples (6 + 25%, 9 + 30%, 12 + 35%, 15%, 18%, and 21 + 50%) and dried for LC-MS/MS analysis.

Six raw files were searched against the six-frame database of M. smegmatis MC² 155 with pFind. The parameters of database searching were as follows: (1) a maximum of three missed cleavages were tolerated; (2) semi-specific cleavage was set; (3) 20 ppm was tolerated in MS and MS/MS search modes; (4) cysteine carbamidomethyl and dimethyl at lysine were set as fixed modification; (5) acetylation of any N-termini, dimethylation at any N-terminal, pyroglutamate at glutamate and glutamine, and oxidation of methionine were added as variable modifications; (6) A minimal peptide length of 7 AAs was required, and (7) the false discovery rates (FDR) for the peptide-spectrum match (PSM); peptide and protein were set to < 1%.

The N-terminal-derived sequences were verified by a comparative genomic approach.

Firstly, the newly translated stop-to-stop protein sequences were compared with NCBI publicly annotated sequences using BLASTP. Secondly, the potential expression at the transcriptional and translational levels was checked by the public RNA-seq and Ribo-seq datasets of M. smegmatis MC² 155, respectively. Thirdly, the potential start codon was confirmed based on the Ribo-seq and direct evidence of N-terminal labeling with dimethylation.

For novel ORFs, novel peptides were assessed by continuous b/y pairs and spectra quality. All potential novel encoding ORFs were rechecked in their annotating state with the publicly available annotated proteins in the NCBI database by BLASTP. The matched orthologous protein sequences were used to construct neighbor-joining (NJ) phylogenetic trees of the novel ORFs encoding proteins by MEGA software (v 10.1.8; Kumar et al., 2018). To observe the conservation of novel ORFs, we used their nucleotide sequences to compare them with 207 reference strains in Mycobacteriaceae by local BLASTN analysis. Sequences sharing at least 60% sequence coverage and 70% identity with other genetically distant species belong to conserved genes and can also be used for reannotation for their homologous genes, as described previously (Gallien et al., 2009).

To achieve deep coverage of proteomics for M. smegmatis MC² 155, we separated two same samples by a 10% SDS-PAGE gel (8 cm) and digested by Ac-Trypsin (Wu et al., 2016) and Ac-LysargiNase (Yang et al., 2019; Zhang et al., 2019) with high activity and stability (Supplementary Figure 1). The extracted peptides were detected by an Orbitrap Fusion Lumos mass spectrometer under the same conditions. The raw files from different proteases were searched against the annotated database for evaluating protein coverage and the six-frame database for selecting novel peptides. To evaluate the contribution of mirror proteases to the accurate identification of novel peptides, we extracted the merging spectra from mirror peptides and calculated the coverage of the b/y ion series. Here, mirror peptides were the overlapped sequences digested simultaneously by Ac-Trypsin and Ac-LysargiNase. In addition, two or more unique peptides from the same novel events were also conserved because of different digesting characteristics. These novel peptides were further assessed by spectra score prediction and their ¹⁵N-labeling spectra matching. All verified novel peptides were used for the genomic reannotation of M. smegmatis MC² 155 (Figure 1).

In Supplementary Figure 2, Ac-Trypsin and Ac-LysargiNase showed higher specificity with more than 95% based on the annotated protein database searching, which also resulted in the identification of more fully digested peptides. In Ac-Trypsin digestion datasets, 47,324 peptides were detected from 4,442 protein groups and 4,102 unique proteins (Supplementary Table 1). In the Ac-LysargiNase datasets, 31,551 peptides were from 3,871 protein groups and 3,602 unique proteins (Supplementary Table 2). The average ratio of the PSM count per peptide and peptides per protein group was more than three and eight (Figure 2A). Among these peptides, 23,175 (40.94%) were mirror peptides, 24,141 unique peptides were from Ac-Trypsin, and 8,748 peptides from Ac-LysargiNase digests (Figure 2B, Supplementary Table 3). At the protein level, 3,478 (78.46%) were shared in two different datasets (Figure 2C), indicating that a larger complement and verification datasets were obtained. The number of non-redundant proteins increased quite steeply and fast with the addition of gel fraction, suggesting the high separation resolution of the gel we used in this study (Figure 2D, Supplementary Figure 1). The total number of unique proteins rose to 4,102 and 3,602 for the Ac-Trypsin and Ac-LysargiNase datasets, which accounted for 64.24 and 56.41% of the total annotated proteins for M. smegmatis MC² 155. In addition, the average protein sequence coverage increased from 27.59% of Ac-LysargiNase and 39.15% of Ac-Trypsin to 42.99% when two separation methods were introduced (Figure 2E, Supplementary Figure 3), indicating a high overlap region of the sequenced peptides and high specificity as well as activity of both proteases we applied in this study.

Due to the cleavage characteristics of proteases, Ac-Trypsin and Ac-LysargiNase digested proteins included significantly more N-terminal and C-terminal peptides, respectively (Figure 2F, P < 0.001; Huesgen et al., 2015). The spectra from these Ac-Trypsin and Ac-LysargiNase digesting peptides had strong y and b series ions, which led to different y and b ion coverages for both Ac-Trypsin and Ac-LysargiNase, respectively. The medium of b and y ion coverages was 29.70 and 74.69% for the Ac-Trypsin spectra, respectively, and 69.37 and 37.01% for the Ac-LysargiNase spectra. In comparison, both y and b ion coverages were improved to >80% when we combined two proteases (Figure 2G, Supplementary Figure 3). We further investigated b and y ion coverages at different positions of all identified peptides. We found that Ac-Trypsin and Ac-LysargiNase spectra could provide complementary ions for whole peptides, including N-terminal, middle, and C-terminal AAs (Supplementary Figure 4). These results support the high quality and deep coverage of M. smegmatis proteome generated by mirror proteases, providing verifiable MS evidence and an opportunity to identify novel peptides.

After the six-frame database search, 845 and 521 novel peptides were identified from Ac-Trypsin and Ac-LysargiNase digests (Figure 3A, Supplementary Tables 4–6). The average ratios for the number of PSMs per novel peptide and novel peptides per protein group were 1.78 and 1.22, which were significantly lower than those of annotated gene products for either Ac-Trypsin or Ac-LysargiNase digested proteome samples. In total, 126 and 119 peptides from 845 Ac-Trypsin and 521 Ac-LysargiNase digests were mirror peptides, respectively. Among them, 111 truncated sequences have the same AAs consistently from Ac-Trypsin and Ac-LysargiNase digests, excepting N- or C-terminal AAs (Supplementary Table 6, Supplementary Figure 5), which resulted in the identification of 74 shared proteins (Figure 3B).

As annotated peptides, Ac-Trypsin and Ac-LysargiNase mirror spectra provided complementary b and y ions with about 80% coverage for novel peptides compared with those of Ac-Trypsin (b-ion, 37.88%; y-ion, 68.25%) or Ac-LysargiNase (b-ion, 65.49%; y-ion, 39.60%; Figure 3C). To further evaluate the accuracy of these novel peptides, we compared the Q-value and raw score of all spectra with single protease and mirror protease evidence. In Figure 3D, the average Q-value of spectra for mirror peptides was 4.91 times lower than that from Ac-Trypsin (single protease, 0.0035; mirror proteases, 0.0007) and 4.22 times lower than that from Ac-LysargiNase (t-test P < 0.0001; single protease, 0.0044; mirror proteases, 0.0010). However, the average raw score of spectra with two proteases digesting evidence was 1.49 times higher than that from Ac-Trypsin (single protease, 6.79; mirror proteases, 10.12) and 1.41 times higher than that from Ac-LysargiNase (t-test P < 0.0001; single protease, 6.46; mirror proteases, 9.14, Figure 3E). These results strongly supported that the novel peptides digested by mirror proteases were more confident than those from the single protease digested samples.

A total of 368 novel mirror peptides were observed and used for proteogenomic analysis. According to the genomic position, these 368 novel peptides were identified from two novel events. Among them, 145 peptides were localized to the N-termini extension regions of 40 annotated genes, and the other 223 peptides from 34 ORFs were in the un-annotating regions of M. smegmatis MC² 155 (Supplementary Table 6, Supplementary Figure 6A). To evaluate the annotation of these novel peptides, we compared their protein sequences to those in the NCBI nr database of other M. smegmatis stains by BLASTP. Results showed that as many as 126 novel peptides from N-termini extension and 185 from novel ORFs were annotated in other annotating versions or the other strains of M. smegmatis, implying that more than 80% of novel peptides were from encoding regions of other M. smegmatis known strains. The remaining 19 and 38 peptides (Supplementary Figures 6B,C) corresponded to six newly calibrated N-terminus and 17 newly identified ORFs (Supplementary Figures 6D,E).

Among the above mentioned 19 N-terminal novel peptides, 12 sequences were mirror peptides, and the other seven peptides were different products digested by mirror proteases (Supplementary Table 7). Among the other 38 novel peptides, 14 sequences belong to mirror peptides, one lysargiNase digested sequence was overlapped with another non-fully tryptic peptide, and the other 22 peptides were different products from mirror proteases digests (Supplementary Table 8). Complementary use of both proteases also increased b/y ion coverage and a sequence coverage of novel events (Supplementary Figures 6B,C). These results imply that mirror proteases provide a higher quality of novel peptides in genome reannotation.

In the 19 novel N-terminal peptides from six annotated proteins identified from the complementary use of both proteases (Figure 4A, Supplementary Table 7, Supplementary Figure 6B), one ORF contained eight novel peptides, 1 ORF contained three peptides, and the other four ORFs each contained two peptides. Among them, 12 novel peptides were detected with essentially identical ¹⁴N and ¹⁵N spectra (cosin >0.9; Supplementary Figure 7). For orf|0|-|626711-627743| and orf|0|+|3349519-3351547|, novel peptides were distributed in front of the start codon GTG (V) and ATG (M) of previously annotated genes. For the other four ORFs, novel peptides covered the area of the originally erroneous start codons. Both mirror N-terminal novel sequences and ¹⁴N/¹⁵N double labeling spectra confirmed the authenticity of the N-termini correction for six recorded genes by Ac-Trypsin and Ac-LysargiNase (Figure 4A).

For example, both peptides (L)SVDTAAPSAVELAGAVTEQLR and (L)SVDTAAPSAVELAGAVTEQL(R) contained as many as 21 AA residues. These two peptides covered the original start codon GTG (V) of the annotated protein WP_011729667.1 for polyprenyl synthetase family protein (Figures 4B,C). As these two peptides are quite long, none of them were sequenced with either full y series or full b series product ions. Complementary with both Ac-Trypsin and Ac-LysargiNase, the sequences of all AA residues were confirmed with at least one product ion. These results proved the correctness and uniqueness of the peptide we identified. In addition to ¹⁴N-labeling spectra, their corresponding ¹⁵N-labeling forms were also identified as high quality. The ¹⁴N and ¹⁵N spectra were essentially the same as the cosin values of similarity (≈1.00) from two peptide pairs generated from Ac-Trypsin and Ac-LysargiNase. Further, we even identified one dimethyl-labeled N-termini peptide started with serine residue with a high-quality spectrum by a dimethylation labeling combined with a negative enrichment strategy (Figure 4D), ^*SVDTAAPSAVELAGAVTEQLR, which included the previously erroneous codon-terminal valine. These results strongly support that the true start codon was CTG for leucine in front of the dimethyl-labeled serine residue considering the characteristics of aminopeptidase (L^*SVDTAAPSAVELAGAVTEQLR, Figure 4D, Supplementary Figure 7).

In addition to mirror peptides, the novel N-terminal events were efficiently confirmed with different peptides identified from the Ac-Trypsin, and Ac-LysargiNase digested proteome. In the case of orf|0|+|3349519-3351547|, we identified two fully digested peptides, (R)GADHDVLR with three continuous b/y pairs and (L)RVQPDVGEL(R) with five b/y pairs, in N-terminal extension region of the recorded protein WP_011728939.1 for ABC transporter ATP-binding protein. Although its N-termini cannot be determined, the true start codon and a potential reason for this mis-annotation are worth elucidating.

In this study, we found erroneous N-termini in six annotated genes. Among them, three were previously annotated at the canonical start codon ATG, whereas the other three were at the non-canonical start codon GTG. Gallien et al. previously found overprediction of ATG and other non-canonical codons as the start codon using a gene prediction program by TMPP labeling strategy in M. smegmatis MC² 155 (Gallien et al., 2009). Kelkar et al. and we also observed similar phenomena in M. tuberculosis (Kelkar et al., 2011; Shi et al., 2022). Other optimized multi-omic technologies, such as Ribo-RET (Meydan et al., 2019; Supplementary Figure 7) and RNA-seq, also provided corresponding expression evidence (Supplementary Figure 8), which further supported our finding through the complementary use of Ac-Trypsin and Ac-LysargiNase for deep coverage proteomics. On the other hand, as shown in Supplementary Figure 7, deep N-terminal proteomics would be the direct evidence for confirming the starting sites of encoding ORFs.

Of the identified 17 novel ORFs with 38 new peptides, six ORFs were detected with mirror peptides, which occupied more than 35% of all novel ORFs identified in this study. Among them, one ORF contained four novel peptides, two ORF contained three peptides, and the other 14 ORFs contained two peptides, indicating the high credibility of these new genes (Figure 5A, Supplementary Table 8). All the novel mirror peptides identified from these novel ORFs were specifically digested products. Among them, four tryptic peptides could also be identified from Blackburn (Potgieter et al., 2016) and Blackburn's (Giddey et al., 2017) datasets after the six-frame database research (Supplementary Table 8, Supplementary Figure 10), further confirming the credibility of these identifications in this study. Surprisingly, we even noticed that orf|0|+|813403-813805| showed quite a high abundance, ranking top 163 of all identified 6,596 genes at the mRNA level (Figure 5B, Supplementary Figure 12). While most of the novel ORFs had much low transcription abundance, which may suggest the deep coverage of the mirror proteome datasets. The reason for the missing annotation remains to be studied.

Mirror proteases also helped identify novel unannotated ORFs with high confidence. For example, four fully digested peptides were derived from the novel protein orf|0|+|1060315-1061149|, including a pair of mirror peptides, (R)DVAQVVGHQHGR and RDVAQVVGHQHG(R), and other two tryptic peptides, (R)DLAGLGVGR and (R)DREHQTDPVDAQR—with at least three continuous b/y ion pairs (Figure 5C).

For mirror peptides, (R)DVAQVVGHQHGR and RDVAQVVGHQHG(R) were identified by 3 and 4 b/y ion pairs in Ac-Trypsin and Ac-LysargiNase digests (Figures 5D,E). However, their meager theoretical peptide, RDVAQVVGHQHGR, five b/y ion pairs were efficiently matched with eight continuous b and y ions by combining Ac-LysargiNase and Ac-Trypsin digestion (Figures 5D,E). In addition, the mirror peptide, RDVAQVVGHQHG(R), was identified with identical ¹⁴N and ¹⁵N spectra (cosin ≈ 1.00). The sequenced AAs were 34 and accounted for 12.2% of the total stop-to-stop translating protein sequence. Most importantly, public Ribo-seq observed the confirmed translation initiation site (TIS) reads (Supplementary Figure 9) at the first leucine in orf|0|+|1060315-1061149| protein sequence, implying CTA (L) might be its start codon. The above evidence not only supported the confident identification but also indicated the N-termini of this novel ORF.

In one extreme case, our datasets detected two different peptides with identical ¹⁴N and ¹⁵N spectra from the novel gene orf|0|+|813403-813805|. The fully tryptic peptide (R)VATPGDSDASAQIEGLR was also identified from the Blackburn (Giddey et al., 2017) dataset, which is the mirror peptide of RVATPGDSDASAQIEGL(R) from our dataset. In addition, two other fully tryptic peptides, (R)LQQESEAFR and (R)WVTVVADALNSASSSGR, were detected in Blackburn's (Potgieter et al., 2016) and Blackburn (Giddey et al., 2017) work (Supplementary Figures 11A–C). Even more importantly, we successfully identified its N-terminal peptide (V)AAAGLAWAVSR with dimethyl modification at the first alanine in our demethylation-labeling N-terminomic dataset (Supplementary Figure 11D) and observed its expression signal at its TIS regions in a public Ribo-seq dataset, implying that the true start codon was GTC (V) in front of an N-terminal peptide and V might be removed by aminopeptidase (Gonzales and Robert-Baudouy, 1996). Thus, we efficiently identified N- and C-terminal peptides of orf|0|+|813403-813805| and confirmed its encoding sequence was 99 AAs. In total, six unique peptides were accurately identified from this novel small ORF, which covered 71.72% of its protein sequence. Together with the high transcription abundance, these results suggest the high abundance of this small novel ORF in M. smegmatis MC² 155 proteome, which might play an important role in the biological process.

Unlike lysargiNase, trypsin-generated N-terminal peptides had C-terminal lysine or arginine residues. We identified 3 N-terminal peptides, (^*)AVDVQHASVAVGARAALGAHR and (^*)AVDVQHASVAVGA(R) from orf|0|-|4966072-4966744| with 224 AAs, and (^*)PLVAPHPGDPAVVVGTLAAVQVEYVR from orf|0|+|1852179-1853043| with 288 AAs. (^*)AVDVQHASVAVGA(R) was an uncompleted mirror peptide of (^*)AVDVQHASVAVGARAALGAHR. These two different N-terminal peptides were derived from Ac-Trypsin, and Ac-LysargiNase digests, respectively. For novel ORF orf|0|+|1852179-1853043|, the other fully digested peptide RLGADEGIGLG(R) was also identified in Ac-LysargiNase digests. This spectra evidence supported the highly confident identification of novel ORFs and indicated the application of mirror proteases in genome re-annotation, even for unannotated ORFs.

To investigate the sequence similarity of 17 novel ORFs, we compared their protein sequences with publicly sequenced and annotated strains by BLASTP in the NCBI reference database. Among these ORFs, 6 ORFs matched other annotated proteins with more than 50% coverage and 40% identity, six ORFs only matched less sequence with lower similarity, and five did not show any homology with recorded proteins in NCBI databases (Supplementary Table 8). Among six higher similarity ORFs, orf|0|+|4368726-4369110| might be a 2-oxoacid dehydrogenases acyltransferase family protein, orf|0|+|3841204-3842056| might be an ion transporter based on the function annotation, while other four ORFs were for hypothetical proteins.

Further BLASTN analysis of these novel ORFs found that 13 ORFs showed higher coverage ranging from 95 to 100% and similarity ranging from 85.71 to 93.57% with their ortholog genes in Mycolicibacterium goodii. Although 13 novel ORFs showed higher homology with those of genetically closer species, they have not been annotated in these stains by far.

The other 4 ORFs showed lower identity with NCBI sequenced strains. The orf|0|-|1217294-1218644| had 100% sequence coverage and 78.06% identity with Nocard ioides okcheonensis MMS20-HV4-12, the orf|0|+|5885997-5886810| had 80% sequence coverage and 67.71% identity with Mycobacterium fortuitum subsp. fortuitum DSM 46621, while orf|0|+|813403-813805| and orf|0|-|5450187-5451333| showed very low coverage and identity with those from the public strains (Supplementary Tables 8, 9, Supplementary Figures 13A–D).

Among 17 novel ORFs, four genes were highly conserved in almost all type strains of 5 genera from the family Mycobacteriaceae, including orf|0|+|3841204-3842056|, orf|0|+|1060315-1061149|, orf|0|+|4368726-4369110|, and orf|0|+|5658318-5659182|, implying that these four novel ORFs may be M. smegmatis species-specific genes.