Cellular cholesterol prediction of mammalian ATP-binding cassette (ABC) proteins based on fuzzy c-means with support vector machine algorithms

Abstract

Over the years protein interaction and prediction of membrane protein have been a pivotal research area for all researchers. For both prokaryotes and eukaryotes Adenosine Triphosphate-(ATP) binding cassette (ABC) genes plays a significant role. In our analysis, we concentrate on human part of ABC genes. In case of living organisms transport of precise molecules across lipid membranes has been treated as vital part and for that reason a bigger transporter is required to carry out the molecules. Here ABC transporter families are evolved to transport the specific molecules such as sugars, amino acid, peptides, proteins, ions etc. within the plasma membrane. As we know another important component of human being is cholesterol, which is a major component in cell membrane and its main functions are to maintain integrity and mechanical stability. Each and every time, membrane cholesterolsareinteracted with membrane protein in both N-C terminuses and target valid sequence(s) which has relevance in human diseases. In this manuscript we have applied Fuzzy C-Means (FCM) with Support Vector Machine (SVM) algorithm for prediction of cellular cholesterol with ABC genes. Our experiments have been performed well using ABCdata set.

Keywords

ABC transporter FCM SVM Prokaryotes/ Eukaryotes CRAC/CARC

1 Introduction

In human pathogenesis cell membranes plays as a major part. Many more proteins are available in cell membranes which have unique functionalities. Among all, ABC proteins are the biggest super family in mammalian membrane. Till now, 49 well characterized ABC genes have been reported. ABC transporter has been utilizing the energy of ATP hydrolysis to translocate solutes across cellular membranes and this protein is responsible for numerous phenomenonssuch as resistance of cancers and pathogenic microbes to drugs. Mammalian ABC transporter familiesare evolved to transport the specific molecules such as sugars, amino acid, peptides, proteins, ions, bile acids etc. within the plasma membrane. On the basis of amino acid homology and other parameter this gene family can be categorized into seven distinct subfamilies like ABCA to ABCG. Cholesterol is a waxy, fat-like substance that is found in every cells of all human being. Basically, cellular cholesterol acts as a significant role in the regulation of membrane proteins [1 –10].

Several membrane cholesterols are binding with proteins, which is the combination of amino acid sequences. In case of human proteins, 20 amino acids have been taken. In order to bind cellular cholesterol with membrane proteins, two different approaches have utilized such as cholesterol recognition/interaction amino acid consensus (CRAC) and reverse of CRAC (CARC). The orientation of CRAC motif is L/V-X(1–5)-Y-X(1–5)-R/K, and CARC is R/K-X(1–5)-Y/F-X(1–5)-L/V where X represents whichever amino acid [11 –15].

This signature motif has been reported in various subfamilies of ABC genes. As each and every time target binding sites may differ according to orientation of protein sequence, therefore it is a tedious task for all researchers to predict the relevant motif sequence which is useful for all drug scientists in recent trends. For this purpose numerous soft computing with data mining techniques such as FCM, SVM, K-Nearest Neighbor (KNN), K-means and Principal Component Analysis (PCA)have been employed by researchers [11 –18].

This paper’s main contribution is to predict cellular cholesterol from ABC genes based on FCM and SVM using protein data and cholesterol motif sequences which would be useful for many drug discovery researchers in future. The structure of the manuscript is as follows: In part 2 the literature works are discussed. Material and proposed models are elaborated in part 3. In part 4 we have detailed description of our methodology and experimental procedure. Lastly we concluded our manuscript in part 5.

2 Literature works

In cell ecology, ABC transporters are well-known as super families of integral membrane proteins and play a part in imperative place in the regulation of growth and resistance processes in human cell membrane. ABC genes are responsible for various human diseases. So, numerous analysts interested to forecast ABC proteins which are useful for clinical pathology. Therefore several prediction techniques are to be obtained in current date.

Hazai, E et al. [19] pursued a support vector machine approach for prediction of breast cancer resistance protein (BCRP) on 164 BCRP substrates and 99 non-substrates data. The advantage of this model was that classification consists of four phases. In first phase, all substrates and non-substrates of wild-type BCRP data were trained and tested then in next phase, compounds in the training set were obtainable as points in a high-dimensional space as they separated the objects into substrate and non-substrate groups based on hyperplane. In third phase, prediction accuracy of the model was computed and finally these proposed models authenticated on the basis of independent external data set. The experimental results specifieds that the SVM model achieves the highest prediction accuracy of 73% than others. In contrast Wang, Z et al. [20] depicted about P-glycoprotein of ABC genes implicated in numerous required procedures like lipid with steroid transport crosswaysplasma membranes. To do so, authors applied a support vector machine based model which consists of 131 substrates and 81 no substrates P-gp data. The total study exhibits that proposed model provide better significance result in separating substrates and nonsubstrates of P-glycoprotein’s. Once again Molinski, S. V et al. [21] proposed biophysical approaches for ABC super family. In their study author indicate that the drug development for diseases caused by dysfunctional ABC proteins.

A phylogenetic analysis is made by Vishwakarma, S. K et al. [22] based on Human ATP Binding Cassette (ABC) Transporters. The proposed model explored the idea on gene superfamily which consists of 49 members of proteins in human and it was seen that phylogenetic analysis method yielded significantly and classified human transporter in a correct manner. Chandra, N [23] established a prediction model using support vector machine algorithm to classify between substrates and nonsubstrates of Pglycoprotein. In their paper authors carried out the research using seven dissimilar types of features along with eight unlike threshold values for each feature type and it includeed features such as unweighted, weighted, Euclidian distance, all pair wise centroid distances among substructures in a compound and all feature combinations. They observed that the model achieved accuracy of 93%. Furthermore Bhavani, S et al. [3] recommended a widespread classifier framework based on support vector machine to predict adverse effects in diverse classes of drugs. Prediction accuracy of the proposed model forecasted accurate results but this method gave more improved output in case of toxicity prediction.

3 Materials and model description

3.1 Data source description

The protein sequences of human ABC transporter were retrieved from uniprot database [24]. Total 49 genes are reported in human ABC super family till date. This family includes many sub family such as ABCA1-ABCA13, ABCB1-ABCB11, ABCC1-ABCC12, ABCG1-ABCG5, ABCG8 etc. Each gene contains protein Id, name of each helix, total length of transmembrane region, accurate position in transmembrane sequence plus its amino acid sequences which are varying in size according with their gene name. Each gene contains dissimilar helix number. Suppose one gene id Q5T3U5 contains helix 1- 17 then other gene id Q96J66 contains 1-10. Maximum 17 numbers of helices are present in ABC transporter. Samples of data set which are retrieved from uniprot database are shown below in Table 1. Another data set of cholesterol dictionary which is constructed using the algorithm CRAC and CARC are shown in Table 2.

Table 1
Sample dataset details containing ABCA1 of ABC superfamily

Genes Protein Id Helix Name Length Position Sequences

ABCA1 O95477 1 21 2-42 LTFRRRQTCQLLLEVAWPLFIFLILI SVRLSYPPY

2 21 640-660 LRVMSRSMPLFMTLAWIYSVAVIIKGIVYEKEARL

3 21 683-703 LDNSILWFSWFISSLIPLLVSAGLLVVILKLGNLL

4 21 716-736 LLPYSDPSVVFVFLSVFAVVTILQCFLISTLFSRA

5 21 745-765 TLFSRANLAAACGGIIYFTLYLPYVLCVAWQDYVG

6 21 777-797 YVGFTLKIFASLLSPVAFGFGCEYFALFEEQGIGV

7 21 827-847 NLTTSVSMMLFDTFLYGVMTWYIEAVFPGQYGIPR

8 21 1041-1057 SGGMQRKLSVALAFVGGSKVVILDEPTAGVD

9 21 1351-1371 RKGFFAQIVLPAVFVCIALVFSLIVPPFGKYPSLE

10 21 1657-1677 TTSVDVLVSICVIFAMSFVPASFVVFLIQERVSKA

11 21 1703-1723 VIYWLSNFVWDMCNYVVPATLVIIIFICFQQKSYV

12 21 1735-1755 SYVSSTNLPVLALLLLLYGWSITPLMYPASFVFKI

13 21 1768-1788 KIPSTAYVVLTSVNLFIGINGSVATFVLELFTDNK

14 21 1802-1822 KLNNINDILKSVFLIFPHFCLGRGLIDMVKNQAMA

15 21 1852-1872 SWDLVGRNLFAMAVEGVVFFLITVLIQYRFFIRPR

Genes	Protein Id	Helix Name	Length	Position	Sequences
ABCA1	O95477	1	21	2-42	LTFRRRQTCQLLLEVAWPLFIFLILI SVRLSYPPY
		2	21	640-660	LRVMSRSMPLFMTLAWIYSVAVIIKGIVYEKEARL
		3	21	683-703	LDNSILWFSWFISSLIPLLVSAGLLVVILKLGNLL
		4	21	716-736	LLPYSDPSVVFVFLSVFAVVTILQCFLISTLFSRA
		5	21	745-765	TLFSRANLAAACGGIIYFTLYLPYVLCVAWQDYVG
		6	21	777-797	YVGFTLKIFASLLSPVAFGFGCEYFALFEEQGIGV
		7	21	827-847	NLTTSVSMMLFDTFLYGVMTWYIEAVFPGQYGIPR
		8	21	1041-1057	SGGMQRKLSVALAFVGGSKVVILDEPTAGVD
		9	21	1351-1371	RKGFFAQIVLPAVFVCIALVFSLIVPPFGKYPSLE
		10	21	1657-1677	TTSVDVLVSICVIFAMSFVPASFVVFLIQERVSKA
		11	21	1703-1723	VIYWLSNFVWDMCNYVVPATLVIIIFICFQQKSYV
		12	21	1735-1755	SYVSSTNLPVLALLLLLYGWSITPLMYPASFVFKI
		13	21	1768-1788	KIPSTAYVVLTSVNLFIGINGSVATFVLELFTDNK
		14	21	1802-1822	KLNNINDILKSVFLIFPHFCLGRGLIDMVKNQAMA
		15	21	1852-1872	SWDLVGRNLFAMAVEGVVFFLITVLIQYRFFIRPR

Table 2

Sample dataset of cholesterol dictionary using algorithms CARC and CRAC

Formula of CRAC and CARC	Motif Type	Cholesterol Length
FORWARD (CRAC)	11, 12,13,14,15	For 11..15 = 5…9, For 21…25 = 6…10
(Leucine/Valine-X_(1-5)-	21, 22,23,24,25	For 31…35= 7…11, For 41…45= 8…12
Tyrosine-X_(1-5)-	31, 32,33,34,35	For 51…55= 9…13
Arginine/Lysine)	41, 42,43,44,45
51, 52,53,54,55
BACKWARD (CARC)	11, 12,13,14,15	For 11..15= 5…9, For 21…25= 6…10
(Arginine/Lysine-X_(1-5)-	21, 22,23,24,25	For 31…35= 7…11,
Tyrosine/Phenylalanine-	31, 32,33,34,35	For 41…45= 8…12
X_(1-5)- Leucine/Valine)	41, 42,43,44,45	For 51…55= 9…13
51, 52,53,54,55

Table 1 represents the sub family of ABC transporter. Here six columns are maintained for gene ABCA1. So many genes are available in uniprot database and with the help of those genes we executed our proposed algorithm. It is not possible to show all the sub genes of the superfamily. Therefore we have taken one sample of gene whose name is ABCA1. Each gene has unique protein Id that is O95477which is shown in table 1. ABCA1 gene has 15 numbers of helices and all helix contain 21 numbers of amino acids. The position of amino acids might be any sequence from transmembrane regions. From transmembrane domain, each helix retrieved their length using different approaches.

Table 2 depicts dictionary construction cholesterol using CRAC (Leucine/Valine-X(1-5)-Tyrosine-X(1-5)-Arginine/Lysine) and CARC (Arginine/Lysine -X(1-5)- Tyrosine /Phenylalanine-X(1-5)- Leucine/Valine) motif. For forward direction, let us assume the motif is LXXYXXK. It means first position is Leucine, second and third positions are any amino acids, then fourth position is Tyrosine, next five and six positions are any amino acid, then last position is Arginine. Here motif type is L2Y2K (22).

Another column represents the length of cholesterol and it varies from 5,6,7,8,9,10,11,12,13.

3.2 Proposed model

In mammalian membrane, proteins are being modulated by cellular cholesterol. As cholesterol plays a pivotal role in cell membrane, most of the researchers interested to do their investigation on cholesterol with membrane protein prediction. Numerous types of membrane proteins are reported current date; amongst them Adenosine Triphosphate Binding Cassette transporters (ABC transporters) has significant role in clinical pathogenesis since it is desired for drug detection for human body.

Figure 1 elucidated the detailed flow of ourplanned model. In phase1 we have created a mammalian cholesterol dictionary using the algorithm CRAC and CARC including window size d5, d6, d7, d8, d9, d10, d11, d12 and d13 etc. Another data set that is ABC genes are also retrieved from uniprot database. After that in phase 2 both data setsare mapped using sliding window concept. Then we extracted both forward and backward motifs in phase 3. In very beginning of the phase 4 we employed a novel hybrid approach FCM with SVM algorithm. As we know FCM algorithm is used to mine such kind of information where data belongs more than one clusters. So in our manuscript we have used FCM and it is well suited for our data set. Then we have taken SVM algorithm embedded with FCM for exact classification among cluster points. At last in phase 5 we identified valid motif sequences of cellular cholesterol which has clinical relevance.

Fig. 1

Schematic outline of proposed model.

4 Methodology and experimental evaluation

4.1 Fuzzy C Means

Cluster study might be represented as assignment of data points to clusters. Things within the identical clusters belong to one cluster while items belonging to diverse clusters are treated as different group of clusters. Furthermore clusters are recognized by resemblance measures. So many clustering algorithms are available for accurate clustering amid the data points. Among all Fuzzy C-means clustering is treated as most important and widespread algorithm in current trends. The main motto of this algorithm is, by means of handing over membership to apiece one data point subsequent to every group center on account of distance among the data points along with cluster center. According to the closeness of data points towards cluster center, the membership has been calculated [25 –27].

Algorithm 1
FCM: Fuzzy C-Means

We can choose the cluster centerarbitrarily;

1. Initialize Z = [membership_ik] matrix, Z⁽⁰⁾

2. Evaluate the membership_ik by means of (3);

3. On j-step: compute the centers vectors center^j = center_k

with Z^(j) by (4).

4. Revise Z^(j), Z^(j +1) with (3);

If ||Z^(j +1)- Z^(j)||< ɛ or the least amount of Cluster center

is got, at that time STOP; or elsego again to step 2.

4.2 Support Vector Machine (SVM)

According to hypothesis, support vector machine is a renowned supervised machine learning approach which is typicallyutilized for resolving classification and regression problems. This algorithm is very simple to implement in current days for solving the class classification problems. SVM concept is based on the decision planes so as to characterize the decision boundaries. A decision plane is one that separates between a set of objects having different class memberships. The foremostintention of this technique is togenerate a line or a hyper plane which divides the data into classes [28 –33].

A training set thatincludes label pairs (u_i, v_i), i = 1, …, m everywhere $u_{i} \in R^{m}$ and v ∈ {profit, loss} ⁱ, SVM algorithm needs outcomebyusing optimization problem that is stated below. $min_{x, y, ξ} \frac{1}{2} z^{N} z + C \sum_{i = 1}^{m} ξ_{i}$ (1)

$Subject to : v_{i} (z^{N} Φ (u_{i}) + y ⩾ 1 - ξ_{i}, ξ_{i} ⩾ 0 .$ (2)

Decision function isspecified below in Equation (3)

$s = pfm (\sum_{i = 1}^{m} x_{i} α_{i} K (x_{i}, x) + ρ)$ (3)

Training vector u_i is mapped in to higher dimensional space by using the kernel function φ. The separating surface depends on a subset of the original data known as a set of support vectors. C > 0 is known as penalty parameter in error case. The theory of SVM tells that, data points are classified on the basis of hyper planes although it becomes unfeasible for getting linear solution in two dimensional spaces. Hence we get rid of this problem by using kernel function w (u_i, u_j) ≡ Φ (u_i) ^NΦ (u_j) for multidimensional data. Utilizingdivergentkernels function, SVM algorithm is trained that is expressed in below Equations (4), (5) plus in Equation (6).

Linear kernel: $(u_{i}, u_{j}) = u_{i}^{N} u_{j}$ (4)

Polynomial kernel: $w (u_{i}, u_{j}) = (γ u_{i}^{N} u_{j} + r)^{e}, γ > 0,$ and (5)

Radial Basis kernel (RBF): (u_i, u_j) = exp(- γ ∥ u_i - u_j ∥ ²) , γ (6)

The entirekernel arguments such as C, γ, r, and e are initialized by utilizing the dataset. All kernel parameters areaffected based upon the size of training data [29 –33].

4.3 Experimental discussion

The proposed algorithm is executed using Intel i3 processor with 4GB hard disk and windows 7 operating system for completing the experiment and total resultant code is written by Python 3. To evaluate the performance of the projected method we have taken ABC gene dataset with cholesterol dictionary. In this paper FCM and SVM are employed for prediction of mammalian cholesterol from ABC proteins. Our proposed works are explained in following phases below.

Phase 1: In first stage we retrieved the protein data information of ABC transporter from uniprot database such as gene, protein Id, helix name which may varies from 1 to 17, length of the protein, position of proteins present in transmembrane region and sequence of protein. Furthermore cholesterol dictionary is constructed using CRAC/CARC algorithm which is described above in data source part.

Phase 2: After preparing both data set then we used sliding window concept for mapping among them. We tookwindow size as D=d5, d6, d7, d8, d9, d10, d11, d12, d13 in this manuscript. According to window size and CRAC/CARC algorithm, cholesterol dictionary is constructed. We know CRAC motif moves on forward direction from N to C terminus. The orientation can be represented as (Leucine/Valine-X_(1 -5)-Tyrosine-X_(1 -5)-Arginine/Lysine). That means Leu/Val takes the first position, followed by part 1 to 5 of any amino acid residue. After that another residue ‘Y’ which is fixed, subsequently once more a part including 1–5 of any residues, and lastly a basic Lys or Arg is normally maintained. Like forward motif, backward motif CARC is also calculated using (Arginine/Lysine -X_(1 -5)- Tyrosine /Phenylalanine-X_(1 -5)- Leucine/Valine).

Phase 3: Once step 2 is finished, we go for next stage where we extracted the motif sequences according with their CRAC/CARC formulas. We divide the motif in to two parts: one for forward direction motif and other is backward motif. Then we move for next phase.

Phase 4: Now our data is well fitted for execution. Here we applied our proposed FCM-SVM algorithm for valid motif prediction.The main motto of this algorithm is by means of handing over membership to apiece one data point subsequent to every group center on account of distance among the data points along with cluster center. According to the closeness of data points towards cluster center, the membership has been calculated.

On the basis of minimization, the algorithm is expressed by means of the below Equations (7) and (8):

$\begin{matrix} {Cluster}_{i} = \sum_{k = 1}^{No . of Motifs} \sum_{j = 1}^{No . of Centers} {membership}_{kj}^{i} \\ {motif}_{k} - {center}_{j} | |^{2}, 1 ⩽ i < \infty \end{matrix}$ (7)

$I_{n} = \sum_{k = 1}^{M} \sum_{j = 1}^{C} v_{kj}^{n} y_{k} - c_{k}^{2}, 1 ⩽ n ⩽ \infty$ (8)

After getting the cluster points concerning membership functions, information is fed to SVM algorithm. For every set of motif pairs on dissimilar positions of ‘Y’ and ‘F’, we found the cluster points. The aim of the SVM is to teach a model that assigns novel hidden objects into a particular group. Now using the cluster points, SVM classifies the motif sequences of cholesterol. Both forward and backward sequences can be observed from these algorithms which have biological relevance in drug design.

Valid motif sequences of cholesterol prediction from ABC genes are shown in Tables 3 and 4. Forward/Backward motifof cholesterol binding with membrane proteins region may produce numerous possibilities. As a result, prediction of cholesterol signature motif was verified for a particular helix according with their sub motifs included in it. A whole amount of CRAC/CARC motifs in a specified ABC proteins are not distributed homogeneously across the membrane, therefore target motif sequences may vary from protein to protein. Most of the proteins those are shown below in Tables 3 and 4 are involved in much functionality like sterol movement from one place to another, accretion or linkedby rafts.From analysis, we concluded that backward motif have more target on proteins than forward motif.

Table 3

Cholesterol motifs derived ABC genes of forward region

Protein Id	Name of Genes	HelicalName	Motif Series	Totality Motif	Begin/LastPosition
ABCA2	Q9BZC7	6,14	VPYMYVAIR	5	Last
			LTIMCQYNFLRR
ABCA5	Q8WWZ7	15	LLQYYEKK	8	Last
ABCA6	Q8N139	7,13	LLLALYFDK	∼	Last
			VLVPSYTLLGFK
ABCA12	Q86UK0	6	VENELSYVLK	2	Last
ABCB2 (TAP1)	Q03518	7	LSLFLWYLVR	3	Last
ABCB3 (TAP2)	Q03519	8	LERALYLLVRR	2	Last
ABCB6	Q9NP58	9	LLCAYFVTEQK	2	Last
ABCB7	O75027	2	VLIGYGVSR	2	Begin
ABCB9	Q9NP78	3	LFVGIYAMVK	2	Begin
ABCB10	Q9NRK6	3	VIYGRYLRK	5	Last
ABCC3	O15438	1	LPCYLLYLR	4	Begin
ABCC4	O15439	6	VAVTLYGAVR	2	Last
ABCC6	O95255	12	VHLAYLR	2	Begin
ABCC8	Q09428	5	VIRVRRYIFFK	2	Last
ABCC9	O60706	14	LGVAFYFIQK	2	Last
ABCC10	Q5T3U5	1	VLSACYLGTPR	2	Last
ABCD3	P28288	4	LATVVGYLVVSR	3	Last
ABCG1	P45844	6	LRLIAYFVLRYK	4	Last
ABCG2	Q9UNQ0	1	LVIGAIYFGLK	2	Last
ABCG4	Q9H172	6	LRLLAYLVLRYR	9	Last

Table 4

Cholesterol motifs derived ABC genes backward region

Protein Id	Name of Genes	Helical Name	Motif Series	Totality Motif	Begin/Last Position
ABCA1	O95477	13	KIPSTAYVVLTSV	3	Begin
ABCA2	Q9BZC7	7	KYFALYEVAGV	5	Begin
ABCA4	P78363	7,11	KDFLAQIVL	2,2	Begin
			RKLLIVFPHFCL
ABCA5	Q8WWZ7	12	KFLAVVFCLIGYV	2	Begin
ABCA7	Q8IZY2	9	RRGLFAQIVL	4	Begin
ABCA8	O94911	3,5	RDSAFWLSWGL	2,6	Begin
			KKSFLTGLVV
ABCA12	Q86UK0	13	KLGAMFVALV	2	Begin
ABCA13	Q86UQ4	10	RMYWFTNFL	2	Begin
ABCB1	P08183	12	KLMSFEDVLLV	4	Begin
ABCB2 (TAP1)	Q03518	5,8,9	RRLSLFLVLVVL	11,9,2	Begin, Last, Begin
			KKVGKWYQLLEV
			KVGILYIGGQLV
ABCB3 (TAP2)	Q03519	8	RALYLLVRRV	3	Begin
ABCB4	P21439	2,7,8	RYAYYYSGL	3,5,2	Begin
			KTEWPYFVVGTV
			KCNIFSLIFL
ABCC1	P33527	3,5,6	RSRGIFLAPVFL	4,4,4	Begin
			RDITFYVYFSLL
			KVLYKTFGPYFL
ABCC2	Q92887	2,6,3	KQVFVGFLLIL	5,7,3	Begin
			KALFKTFYMVLL
			RAPAPVFFVTPLV
ABCC3	O15438	5	RFTTFYIHFAL	3	Begin
ABCC4	O15439	9	RSLLVFYVLV	5	Begin
ABCC5	O15440	6	KVTPFSVKSL	2	Last
ABCC6	O95255	2,7,14	KMVLGFALIVL	3,2,2	Begin
			KGYLLAVL
			RSLLMYAFGLL
ABCC8	Q09428	16	RMEYIGACVV	2	Begin
ABCC9	O60706	10,11	KTFALYTSL	3,2	Begin
			KPAEAFASLSL
ABCC11	Q96J66	6	RLSVFFVPIAV	3	Last
ABCC12	Q96J65	7	KASGGYLLSL	2	Last
ABCD1	P33897	1,2	RTFLSVYVARL	3,6	Begin/Last
			RKDPRAFGWQLL
ABCD3	P28288	2	KRYLLNFIAAMPL	10	Begin
ABCG1	P45844	6	KLYLDFIVL	7	Begin
ABCG2	Q9UNQ0	4	KPKADAFFV	3	Begin
ABCG5	Q9H222	4	RFGYFSAALL	4	Begin

We can assume that more number of sub motifs we found from investigation have better interactionwith membrane cholesterol. Overall, the studies saythat enhancement of the sub motifs’ target region with membrane cholesterol which increased the predictability of the cellular cholesterol consensus motif.

5 Conclusion

In case of human physiology research, on daily basis new problems and facts are created. For solving such type of biological issues many soft computing algorithms has been implemented by researchers. Quite a few membrane proteins like ABC proteins, GPCR proteins are reported to be modulated by membrane cholesterol. In case of living organisms transport of precise molecules across lipid membranes are treated as vital part and for that reason a bigger transporter is required to carry out the molecules. So our current study emphasized on ABC transporter with cellular cholesterol where the transmembrane-sequence matching of forward/backward cholesterol motif is extracted. In this manuscript we applied Fuzzy C-Means with Support Vector Machine algorithm for prediction of cellular cholesterol with ABC genes. Here we retrieved the protein data information of ABC transporter from uniprot database. Furthermore cholesterol dictionary has been constructed using CRAC/CARC algorithm which is described above in data source part. Our experiments is performed well using the given data set and much more valid motif of different amino acid sequences are found which has clinical relevance in human being for multi-drug discovery.

References

Higgins

Christopher F.

, ABC transporters: from microorganisms to man, Annual Review of Cell Biology8(1) (1992), 67–113.

Hyde

Stephen C.

, Emsley

Paul

, Hartshorn

Michael J.

, Mimmack

Michael M.

, Gileadi,

Uzi

, Pearce,

Stephen R.

, Gallagher,

Maurice P.

, Gill,

Deborah R.

, Hubbard

Roderick E.

and Higgins

Christopher F.

, Structural model of ATP-binding proteing associated with cystic fibrosis, multidrug resistance and bacterial transport, Nature346(6282) (1990), 362.

Paila

Yamuna Devi

and Chattopadhyay

Amitabha

, Membrane cholesterol in the function and organization of G-protein coupled receptors, In Cholesterol Binding and Cholesterol Transport Proteins:, pp. 439–466. Springer, Dordrecht, 2010.

Schmitz

Gerd

and Kaminski

Wolfgang E.

, Phospholipid transporters ABCA1 and ABCA7. In Membrane Transporter Diseases, pp. 291–299. Springer, Boston, MA, 2003.

Tripathy

Ramamani

, Mishra

Debahuti

and BadireenathKonkimalla

, A novel fuzzy C-means approach for uncovering cholesterol consensus motif from human G-protein coupled receptors (GPCR), Karbala International Journal of Modern Science1(4) (2015), 212–224.

Walker

John E.

, Saraste

Matti

, Runswick

Michael J.

and Gay

Nicholas J.

, Distantly related sequences in the alpha-and beta-subunits of ATP synthase, myosin, kinases and other ATP-requiring enzymes and a common nucleotide binding fold, The EMBO Journal1(8) (1982), 945–951.

Coyle

Beth

, Kessler

Maya

, Sabnis

Durgagauri H.

and Kerr

Ian D.

, ABCB1 in children’s brain tumours, Biochemical Society Transactions43(5) (2015), 1018–1022.

Baier

Carlos J.

, Fantini

Jacques

and Barrantes

Francisco J.

, Disclosure of cholesterol recognition motifs in transmembrane domains of the human nicotinic acetylcholine receptor, Scientific Reports1 (2011), 69.

Pucadyil

Thomas J.

and Chattopadhyay

Amitabha

, Role of cholesterol in the function and organization of G-protein coupled receptors, Progress in Lipid Research45(4) (2006), 295–333.

10.

Tripathy

Ramamani

, Mishra

Debahuti

, BadireenathKonkimalla

and Nayak

RudraKalyan

, A computational approach for mining cholesterol and their potential target against GPCR seven helices based on spectral clustering and fuzzy c-means algorithms, Journal of Intelligent & Fuzzy Systems Preprint (2018), 1–10.

11.

Yang

Zhang

, DeVries

Mark E.

and Skolnick

Jeffrey

, Structure modeling of all identified G protein–coupled receptors in the human genome, PLoS Computational Biology2(2) (2006), e13.

12.

Doeven

Mark K.

, Biemans-Oldehinkel

Esther

and Poolman

Bert

, ABC transporter architecture and regulatory roles of accessory domains.

13.

Patel

Dipesh C.

, Albrecht

Christiane

, Pavitt

Darrell

, Paul

Vijay

, Pourreyron

Celine

, Newman

Simon P.

, Godsland

Ian F.

, Valabhji

Jonathan

and Johnston

Desmond G.

, Type 2 diabetes is associated with reduced ATP-binding cassette transporter A1 gene expression, protein and function, PLoS One6(7) (2011), e22142.

14.

Brooks-Wilson

Angela

, Marcil

Michel

, Clee

Susanne M.

, Zhang

Lin-Hua

, Roomp

Kirsten

, van Dam

Marjel

, Yu

, et al., Mutations in ABC1 in Tangier disease and familial high-density lipoprotein deficiency, Nature Genetics22(4) (1999), 336.

15.

Lawn

Richard M.

, Wade

David P.

, Garvin

Michael R.

, Wang

Xingbo

, Schwartz

Karen

, Gordon Porter

, Seilhamer

Jeffrey J.

, Vaughan

Ashley M.

and Oram

John F.

, The Tangier disease gene product ABC1 controls the cellular apolipoprotein-mediated lipid removal pathway, The Journal of Clinical Investigation104(8) (1999), R25–R31.

16.

Mendez

Armando J

, Cholesterol efflux mediated by apolipoproteins is an active cellular process distinct from efflux mediated by passive diffusion, Journal of Lipid Research38(9) (1997), 1807–1821.

17.

Oram

John F.

, Mendez

Armando J.

, Lymp

James

, Kavanagh

Terrance J.

and Halbert

Christine L.

, Reduction in apolipoprotein-mediated removal of cellular lipids by immortalization of human fibroblasts and its reversion by cAMP: lack of effect with Tangier disease cells, Journal of Lipid Research40(10) (1999), 1769–1781.

18.

Gottesman

Michael M.

, Fojo

Tito

and Bates

Susan E.

, Multidrug resistance in cancer: role of ATP–dependent transporters, Nature Reviews Cancer2(1) (2002), 48.

19.

Hazai

Eszter

, Hazai

Istvan

, Ragueneau-Majlessi

Isabelle

, Chung

Sophie P.

, Bikadi

Zsolt

and Mao

Qingcheng

, Predicting substrates of the human breast cancer resistance protein using a support vector machine method, BMC Bioinformatics14(1) (2013), 130.

20.

Wang

Zhi

, Yuanying

Chen

, Liang andreas Bender

, Glen

Robert C.

and Yan

Aixia

, P-glycoprotein substrate models using support vector machines based on a comprehensive data set, Journal of Chemical Information and Modeling51(6) (2011), 1447–1456.

21.

Molinski

Steven V.

, Bozóky

Zoltán

, Iram

Surtaj H.

and Ahmadi

Saumel

, Biophysical approaches facilitate computational drug discovery for ATP-binding cassette proteins, International Journal of Medicinal Chemistry2017 (2017).

22.

Vishwakarma

Sandeep Kumar

, Ameer

, Paspala

and Khan

A. A.

, Human ATP Binding Cassette (ABC) Transporters: A Phylogenetic Investigation, Int J Sci Res3 (2014), 564–571.

23.

Chandra

Nagasuma

, Support Vector Machine Classifier for Predicting Drug Binding to P-glycoprotein, (2009).

24.

UniProt Consortium, The universal protein resource (UniProt), Nucleic Acids Research36(1) (2007), D190–D195.

25.

Ahmed

Mohamed N.

, Yamany

Sameh M.

, Mohamed

Nevin

, Farag

Aly A.

and Moriarty

Thomas

, A modified fuzzy c-means algorithm for bias field estimation and segmentation of MRI data, IEEE Transactions on Medical Imaging21(3) (2002), 193–199.

26.

Sreenivasarao

Vuda

and Vidyavathi

Dr S.

, Comparative analysis of fuzzy C-mean and modified fuzzy possibilistic C-mean algorithms in data mining, IJCST1(1) (2010), 104–106.

27.

Lin

Phen-Lan

, Huang

Po-Whei

, Kuo

Chun-Hung

and Lai

Y. H.

, A size-insensitive integrity-based fuzzy c-means method for data clustering, Pattern Recognition47(5) (2014), 2042–2056.

28.

Nayak

RudraKalyan

, Mishra

Debahuti

and Kumar Rath

Amiya

, A Naïve SVM-KNN based stock market trend reversal analysis for Indian benchmark indices, Applied Soft Computing35 (2015), 670–680.

29.

Cortes

Corinna

, WSupport-vector network, Machine Learning20 (1995), 1–25.

30.

Tay

Francis EH

and Cao

Lijuan

, Application of support vector machines in financial time series forecasting, Omega29(4) (2001), 309–317.

31.

Nayak

RudraKalyan

, Mishra

Debahuti

and Kumar Rath

Amiya

, An optimized SVM-k-NN currency exchange forecasting model for Indian currency market, Neural Computing and Applications31(7) (2019), 2995–3021.

32.

Zhang

Xueying

and Xianghan

Zheng

, Comparison of Text Sentiment Analysis Based on Machine Learning, In 2016 15th International Symposium on Parallel and Distributed Computing (ISPDC), pp. 230–233. IEEE, 2016.

33.

Upadhyay

VedPrakash

, Panwar

Subhash

, Ramchander-Merugu and Panchariya

Ravindra

, Forecasting stock market movements using various kernel functions in support vector machine. In Proceedings of the International Conference on Advances in Information Communication Technology & Computing, p. 107. ACM, 2016.