Machine Learning Discoveries of ATG3-X Synergy in ETC-1922159 Treated Colorectal Cancer Cells

1. Introduction

1.1. Autophagy

Authophagy is a natural process in which cells works towards degradation of dysfunctional proteins and other cytoplasmic cargo via lysosome dependent mechanism. Hitherto, there are three ways by which autophagy happens - (1) macroautopahgy, (2) chaperone mediated autophagy and (3) microautophagy. Out of these the first one is the most widely researched topic. Autophagy was coined by Christian de Duve in 1963, after his discovery of the exitence of lysosomes which were involved in the process (De Duve and Wattiaux [1] and Klionsky [2]). Later, Takeshige et al. [3] first observed the autophagy degradation in yeast cells. This was followed by isolation and characterization of autophagy causing genes in Tsukada and Ohsumi [4]. Introductory reviews on autophagy can be found in Levine and Kroemer [5] and Levine and Kroemer [6]. The reviews also cover the roles of autophagy genes in cancer, briefly. A recent study by Li et al. [7], discusses the role of autophagy and its related genes (i.e ATGs) in both cancer suppression as well as cancer promotion.

1.2. ATG3

ATG3 is a protein that lacks rigid structure and more specifically it is a ubiquitin carrier protein E2-like enzyme. The crystal structure of ATG3 has been elucidated in Yamada et al. [8]. ATG3 engages with many binding partners and binding sites. Fang et al. [9] indicate that ATG3 interacts with ATG7 via formation of an E1-E2 complex, LC3/ATG8 via a thioester bond and ATG12 via leucine of the LC3-interacting region motif, while providing necessary references for the same. They also cite references which show that ATG3 is implicated in various types of cancers. In a recent finding, Huang et al. [10] show that ATG3 promotes colon cancer. ATG3 was found to be down regulated in colorectal cancer cell lines after the treatment of ETC-1922159 drug as observed in Madan et al. [11]. Most studies till now, have dealt with how the ATG3 works and there is very less informantion reagarding which gene/protein combinations might be working synergistically along with it. It would be nice to observe if there is any connection between the independently observed factors in the form of unknown biological hypotheses. To solve the issue, the next section discusses a solution to the problem.

1.3. Combinatorial search problem and a possible solution

In a recently published work Sinha [12], a frame work of a search engine was developed which can rank combinations of factors (genes/proteins) in a signaling pathway. Readers are requested to go through the adaptation of the above mentioned work for gaining deeper insight into the working of the pipeline and its use of published data set generated after administration of ETC-1922159, Sinha [13]. The work uses SVM package by Joachims [14] in https://www.cs.cornell.edu/people/tj/svm_light/svm_rank.html. I use the adaptation to rank 2^nd order gene combinations.

2. Results & Discussion

2.1. ATG3 related synergies

2.1.1. ATG3-PDCD

Murrow et al. [15] identify an interaction between ATG12-ATG3 and PDCD6IP and demonstrate that the interaction controls multiple PDCD6IP dependent processes like exosome biogenesis, late endosome distribution,, and viral budding. In colorectal cancer cells treated with ETC-1922159, PDCD family members and ATG3, were found to be down regulated and recorded independently. I was able to rank 2^nd order combinations of PDCD family members and ATG3, that were down regulated.

Table 1 shows rankings of these combinations. Followed by this is the unexplored combinatorial hypotheses in Table 2 generated from analysis of the ranks in Table 1. The Table 1 shows rankings of PDCD family w.r.t ATG3. PDCD2 - ATG3 shows low ranking of 338 (laplace) and 272 (linear). These rankings point to the synergy existing between the two components, which have been down regulated after the drug treatment.

Further, PDCD2L, PDCD11 and PDCD7 showed high ranking with ATG3, thus indicating that they might not be working synergistically with ATG3, before the drug treatment.

One can also interpret the results of the Table 1 graphically, with the following influences - • PDCD family w.r.t ATG3 with ATG3 $->$ PDCD2.

2.1.2. ATG3-BRD

In acute myelogenous leukemia cells, Huang et al. [16] found that BRD4 binds to the promoters of ATG3, and expression of this gene is reduced by inhibitors of BRD4. Thus BRD4 plays a direct role in autophagy by regulating the transcription of ATG3. In colorectal cancer cells treated with ETC-1922159, BRD family members and ATG3, were found to be down regulated and recorded independently. I was able to rank 2^nd order combinations of BRD family members and ATG3, that were down regulated.

Table 3 shows rankings of these combinations. Followed by this is the unexplored combinatorial hypotheses in Table 4 generated from analysis of the ranks in Table 3. The Table 3 shows rankings of BRD family w.r.t ATG3. BRD8 - ATG3 shows low ranking of 453 (laplace), 574 (linear) and 1550 (rbf). These rankings point to the synergy existing between the two components, which have been down regulated after the drug treatment.

One can also interpret the results of the Table 3 graphically, with the following influences - • BRD family w.r.t ATG3 with ATG3 $->$ BRD8.

2.1.3. ATG3-DARS/METTL

In cervical cancer (CC), Shen et al. [17] experimentally confirmed that DARS-AS1 regulated the expression of ATG3 to affect CC cell autophagy by modulating DARS expression. Further, they show that DARS-AS1 recruits METTL3 and METTL14 mediated m⁶A methylation to translate DARS translation. In colorectal cancer cells treated with ETC-1922159, DARS (and METTL) family members and ATG3, were found to be down regulated and recorded independently. I was able to rank 2^nd order combinations of DARS (and METTL) family members and ATG3, that were down regulated.

Table 5 shows rankings of these combinations. Followed by this is the unexplored combinatorial hypotheses in Table 6 generated from analysis of the ranks in Table 5. The Table 5 shows rankings of DARS family w.r.t ATG3. DARS2 - ATG3 shows low ranking of 1242 (laplace), 1451 (linear) and 512 (rbf). METTL1 - ATG3 shows low ranking of 472 (laplace), 337 (linear) and 391 (rbf). METTL8 - ATG3 shows low ranking of 1287 (laplace) and 1469 (rbf). These rankings point to the synergy existing between the two components, which have been down regulated after the drug treatment.

Further, METTL13, METTL3, METTL21A, METTL17, METTL16, METTL5, METTL12, METTL21B and METTL2B showed high ranking with ATG3, thus indicating that they might not be working synergistically with ATG3, before the drug treatment.

One can also interpret the results of the Table 5 graphically, with the following influences - • DARS family w.r.t ATG3 with ATG3 $->$ DARS2 and ATG3 $->$ METTL-1/8.

2.1.4. ATG3-EIF

EIF5A function is well-described in yeast and bacteria, but little is known about its translational targets in human cells. Frankel [18] using liquid chromatography-mass spectrometry (LC-MS) analysis, revealed that EIF5A affects the translation of ATG3. It was confirmed that ATG3 protein levels were reduced upon knockdown of EIF5A. In colorectal cancer cells treated with ETC-1922159, EIF family members and ATG3, were found to be down regulated and recorded independently. I was able to rank 2^nd order combinations of EIF family members and ATG3, that were down regulated.

Table 7 shows rankings of these combinations. Followed by this is the unexplored combinatorial hypotheses in Table 8 generated from analysis of the ranks in Table 7. The Table 7 shows rankings of EIF family w.r.t ATG3. EIF2B1 - ATG3 shows low ranking of 784 (laplace) and 485 (linear). EIF2D - ATG3 shows low ranking of 1300 (laplace), 1108 (linear) and 1497 (rbf). These rankings point to the synergy existing between the two components, which have been down regulated after the drug treatment.

Further, EIF2B3, EIF3F, EIF3L, EIF3E, EIF2AK4, EIF4B, EIF2B5 and EIF4EBP1 showed high ranking with ATG3, thus indicating that they might not be working synergistically with ATG3, before the drug treatment.

One can also interpret the results of the Table 7 graphically, with the following influences - • EIF family w.r.t ATG3 with ATG3 $->$ EIF-2B1/2D.

2.1.5. ATG3-FOX

In hypoxic granulosa cells, Li et al. [19] show that members of forkhead box proteins FOXO, stimulate the upregulation of ATG3. In colorectal cancer cells treated with ETC-1922159, FOX family members and ATG3, were found to be down regulated and recorded independently. I was able to rank 2^nd order combinations of FOX family members and ATG3, that were down regulated.

Table 9 shows rankings of these combinations. Followed by this is the unexplored combinatorial hypotheses in Table 10 generated from analysis of the ranks in Table 9. The Table 9 shows rankings of FOX family w.r.t ATG3. FOXM1 - ATG3 shows low ranking of 14 (laplace), 66 (linear) and 147 (rbf). These rankings point to the synergy existing between the two components, which have been down regulated after the drug treatment.

Further, FOXD2-AS1, FOXA2 and FOXJ1 showed high ranking with ATG3, thus indicating that they might not be working synergistically with ATG3, before the drug treatment.

One can also interpret the results of the Table 9 graphically, with the following influences - • FOX family w.r.t ATG3 with ATG3 $->$ FOXM1.

2.1.6. ATG3-GAS

Li et al. [20] showed that knockdown of GAS5 suppressed the expression of LC3II, ATG3 and ATG5-ATG12 complex formation, thus suggesting that GAS5/miR-23a/ATG3 axis might be a regulatory network contributing to autophagy and cell viability. In colorectal cancer cells treated with ETC-1922159, GAS family members and ATG3, were found to be down regulated and recorded independently. I was able to rank 2^nd order combinations of GAS family members and ATG3, that were down regulated.

The Table 11 shows rankings of GAS family w.r.t ATG3. Interestingly, all recorded family members, i.e GAS5, GAS5-AS1, GAS2L3, GAS6 and GAS6-AS1 showed high ranking with ATG3, thus indicating that they might not be working synergistically with ATG3, before the drug treatment.

2.1.7. ATG3-RAB

ATG16L facilitates LC3/ATG8-conjugation to phos-phatidylethanolamine by forming a complex with ATG12-conjugated ATG5 and recruiting an LC3-ATG3 intermediate to elongating isolation membranes. Fukuda and Itoh [21] report that ATG16L interacts with the Golgi-resident small GTPase RAB33B and RAB33A. Thus there exists a synergy or connection between RAB33-A/B and ATG3. In colorectal cancer cells treated with ETC-1922159, RAB family members and ATG3, were found to be down regulated and recorded independently. I was able to rank 2^nd order combinations of RAB family members and ATG3, that were down regulated.

Table 12 shows rankings of these combinations. Followed by this is the unexplored combinatorial hypotheses in Table 13 generated from analysis of the ranks in Table 12. The Table 12 shows rankings of RAB family w.r.t ATG3. RABEPK - ATG3 shows low ranking of 634 (laplace) and 1261 (rbf). RAB26 - ATG3 shows low ranking of 851 (laplace), 1180 (linear) and 157 (rbf). RAB23 - ATG3 shows low ranking of 1121 (laplace) and 1073 (rbf). These rankings point to the synergy existing between the two components, which have been down regulated after the drug treatment.

Further, RAB40B, RAB11FIP3, RABL5, RAB36 and RAB40A showed high ranking with ATG3, thus indicating that they might not be working synergistically with ATG3, before the drug treatment.

One can also interpret the results of the Table 12 graphically, with the following influences - • RAB family w.r.t ATG3 with ATG3 $->$ RAB-EPK/26/23.

2.1.8. ATG3-RAD51

RAD51 plays a major role in homologous recombination but it is unclear whether RAD51 can be involved in gene regulation as a co-factor. Kang et al. [22] show results which suggest that RAD51 contributes to the regulation of autophagy-related genes like ATG3 and ATG5 in a DNA-binding-dependent manner. In colorectal cancer cells treated with ETC-1922159, RAD51 family members and ATG3, were found to be down regulated and recorded independently. I was able to rank 2^nd order combinations of RAD51 family members and ATG3, that were down regulated.

Table 14 shows rankings of these combinations. Followed by this is the unexplored combinatorial hypotheses in Table 15 generated from analysis of the ranks in Table 14. The Table 14 shows rankings of RAD51 family w.r.t ATG3. RAD51 - ATG3 shows low ranking of 142 (laplace), 1039 (linear) and 642 (rbf). RAD51C - ATG3 shows low ranking of 757 (laplace), 1200 (linear) and 291 (rbf). RAD51AP1 - ATG3 shows low ranking of 67 (laplace), 644 (linear) and 385 (rbf). These rankings point to the synergy existing between the two components, which have been down regulated after the drug treatment.

One can also interpret the results of the Table 14 graphically, with the following influences - • RAD51 family w.r.t ATG3 with ATG3 $->$ RAD-51/51C/51AP1.

2.1.9. ATG3-TRIM

After L. monocytogenes infection, Wang et al. [23] found that TRIM7 overexpression resulted in enhanced LC3-ATG3 association, thus implicating TRIM7 as a regulator for autophagy. In colorectal cancer cells treated with ETC-1922159, TRIM family members and ATG3, were found to be down regulated and recorded independently. I was able to rank 2^nd order combinations of TRIM family members and ATG3, that were down regulated.

Table 16 shows rankings of these combinations. Followed by this is the unexplored combinatorial hypotheses in Table 17 generated from analysis of the ranks in Table 16. The Table 16 shows rankings of TRIM family w.r.t ATG3. TRIM59 - ATG3 shows low ranking of 1061 (laplace), 458 (linear) and 317 (rbf). TRIM28 - ATG3 shows low ranking of 1351 (laplace) and 727 (linear). These rankings point to the synergy existing between the two components, which have been down regulated after the drug treatment.

Further, TRIM32, TRIM7 and TRIM65 showed high ranking with ATG3, thus indicating that they might not be working synergistically with ATG3, before the drug treatment.

One can also interpret the results of the Table 16 graphically, with the following influences - • TRIM family w.r.t ATG3 with ATG3 $->$ TRIM-59/28.

2.1.10. ATG3-TNFAIP

TNFAIP8 regulates autophagy by interacting with ATG3-ATG7 autophagosome complex proteins and promotes hepatocellular carcinoma cell proliferation, as shown by Niture et al. [24]. Thus these is a direct connection between TNFAIP8 and ATG3. In colorectal cancer cells treated with ETC-1922159, TNFAIP family members and ATG3, were found to be down regulated and recorded independently. I was able to rank 2^nd order combinations of TNFAIP family members and ATG3, that were down regulated.

Table 18 shows rankings of these combinations. Followed by this is the unexplored combinatorial hypotheses in Table 19 generated from analysis of the ranks in Table 18. The Table 18 shows rankings of TNFAIP family w.r.t ATG3. TNFAIP8L1 - ATG3 shows low ranking of 781 (laplace) and 126 (rbf). These rankings point to the synergy existing between the two components, which have been down regulated after the drug treatment.

One can also interpret the results of the Table 18 graphically, with the following influences - • TNFAIP family w.r.t ATG3 with ATG3 $->$ TNFAIP8L1.

3. Conclusion

Presented here are a range of multiple synergistic ATG3 2^nd order combinations that were ranked via a machine learning based search engine. Via majority voting across the ranking methods, it was possible to find plausible unexplored synergistic combinations of ATG3-X that might be prevalent in CRC cells after treatment with ETC-1922159 drug.