Novel Descriptors Of Amino Acids And Their Applications In Peptide QSAR

Peptides play a signifcant role in biological world especially in human lifeand participate in a vast array of biological functions. Therefore, peptides are infocus of innovative drug development efforts backed up by many chemists,biologists and medical scientists. Major information of structure and function forpeptides is contained in their amino acid sequences. Therefore, characteristics ofthe amino acid residues are of great significance to QSAR study of peptides. Twokinds of amino acid descriptors, i.e.principal component score vector of radialdistribution function descriptors and geometrical descriptors （SVRG） andprincipal component score vector of electronic eigenvalue descriptors （SVEEVA）,were extracted from principal component analysis （PCA）. SVRG was derivedfrom PCA of independent families of150RDF descriptors and74geometricaldescriptors, respectively, which were in total224steric parameters of20codedamino acids. With regard to each amino acid, SVRG1～SVRG10are related toRDF properties, SVRG₁₁～SVRG₁₆indicate geometrical properties. SVEEVAwas derived from PCA of220EEVA variables of20coded amino acids. Theresults showed that both SVRG and SVEEVA scales have many distinctcharacteristics such as high characterization competence, easy manipulation andconvenient expansibility.SVRG and SVEEVA were then applied to describe the chemical structures ofseveral functional peptides, including angiotensin-converting enzyme inhibitors,bitter tasting threshold of dipeptide, oxytocin analogues and antimicrobialpeptides. Here some quantitative structure activity relationship （QSAR） modelswere built by stepwise linear regression-partial least square regression（SMR-PLS） or stepwise linear regression-multiple linear regression （SMR-MLR）.The estimation stability and predictive power of the models were strictlyanalyzed by internal validation and external validation. The results showed thatthe novel amino acid descriptors could preferably express structure information of peptides, which had both favorable estimation stability and good predictioncapabilities.In the first part of the dissertation, not noly a brief description of QSAR,peptide QSAR development process and research status was given, and alsoexpatiate on the study objective and significance of this dissertation.In the second part of the dissertation, both the structural description andresearch methods in peptide QSAR study were outlined. Structural description isa key step in the QSAR studies. Two kinds of amino acid descriptors, i.e. SVRGand SVEEVA, were derived. The modeling methods and related techniques arealso important for the success of QSAR studies. Multiple linear regression（MLR）, principal component analysis （PCA） and partial least squares （PLS） weresummarized in this chapter. In addition, feature selection and model validationwere summarized in this chapter.In the third part of the dissertation, based on the intense researches on twokinds of novel amino acid descriptors, i.e. SVRG and SVEEVA, the QSARstudies related to58angiotensin-converting enzyme inhibitors,48bitter tastingdipeptides,21oxytocin analogues, and12antimicrobial peptides were dwelledon in detail.The last part of this thesis is the summary of the study. Moreover, aconclusion about the further development and ideas based on the summarizationswas given.