CN106951735A - A kind of Forecasting Methodology of signal peptide and its cleavage site based on layering mixed model - Google Patents
A kind of Forecasting Methodology of signal peptide and its cleavage site based on layering mixed model Download PDFInfo
- Publication number
- CN106951735A CN106951735A CN201710140273.XA CN201710140273A CN106951735A CN 106951735 A CN106951735 A CN 106951735A CN 201710140273 A CN201710140273 A CN 201710140273A CN 106951735 A CN106951735 A CN 106951735A
- Authority
- CN
- China
- Prior art keywords
- signal peptide
- sequence
- amino acid
- cleavage site
- fraction
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
- G16B40/00—ICT specially adapted for biostatistics; ICT specially adapted for bioinformatics-related machine learning or data mining, e.g. knowledge discovery or pattern finding
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
- G16B30/00—ICT specially adapted for sequence analysis involving nucleotides or amino acids
Abstract
The invention discloses the Forecasting Methodology of a kind of signal peptide based on layering mixed model and its cleavage site, whether contain N-terminal hydrophobic patches using the SVM classifier identification of protein sequence based on amino acid residue feature in first layer first；Then it is signal peptide or N-terminal transbilayer helix to be applied in the second layer based on amino acid residue feature and the corresponding naive Bayesian of functional structure characteristic of field and SVM classifier identification hydrophobic patches；Finally in third layer, first according to statistical learning Rules Filtering candidate cleavage site and counting statistics trust fraction, then the similarity scores of candidate signal peptide sequence are calculated by Needleman Wunsch sequence alignment algorithms again, to the statistical signal peptide cutting site for trusting fraction and the integration determination prediction of sequence similarity fraction.
Description
Technical field
It is using the present invention relates to the Forecasting Methodology of a kind of signal peptide based on layering mixed model and its cleavage site
The protein sequence known predicts that whether the protein includes N-terminal signal peptide, and predict its cleavage site, particularly a kind of fusion
Amino acid residue and functional domain, fusion is statistical to trust fraction and sequence similarity fraction, and lower leaf is predicted from top
The algorithm of signal peptide and its cleavage site.
Background technology
1979, G.Blobel and D.Sabatini proposed signal hypothesis (Signal first based on Germicidal efficacy
hypothesis).G.Blobel and D.Sabatini think that the N-terminal in secreted protein sequence has one section to play signal guiding function
Amino acid fragment, the fragment can with pilot protein matter each it is intermembranous shifted, and by protein import to destination locations.
The amino acid fragment that this section plays signal guiding function is called signal peptide by they.In 1999, G.Blobel was achieved by this
Obtain Nobel prize's soul.
Research by years of researches to signal peptide, people further deepen to the understanding of signal peptide.Signal peptide is one
Section appears in the short sequence of amino acid of protein sequence N-terminal, it in eucaryote and prokaryotes pilot protein matter in cell
In transfer and secretion.The N-terminal of almost all of secretory protein and many transmembrane protein amino acid sequences all includes signal
Peptide.As one " postcode ", signal peptide guiding nascent protein transports the correct position of intraor extracellular.If changing new life
The signal peptide of albumen n end, then protein may be transported to the cell position of mistake, and this will cause various strange diseases
Disease.Therefore, the knowledge of signal peptide is highly useful to the mechanism for disclosing some complex inheritance diseases.Further according to signal
The knowledge of peptide, scientist can be reprogramed in oneself desired mode to amino acid sequence, and this can be used for the thin of future
In born of the same parents and gene therapy, and provide thinking simultaneously for developing new drug thing.In fact, pharmaceutical science man passes through genetic modification albumen
Confrontation induction signal peptide, for example, scientist be able to can be marked simultaneously by adding special signature's (signal peptide) to required protein
Guide their excretion so that they are easier to be acquired.Now, signal peptide turns into finds novel drugs, and cell is weighed
New program and the key tool for carrying out gene therapy.In addition, signal peptide prediction can also provide for the prediction of other association areas
Important reference information, such as in Subcellular Localization research, the information of signal peptide can directly provide the subcellular fraction position of protein
Confidence ceases.
Over the last couple of decades, science is always in the tireless effort of signal peptide prediction and its Locus Analysis in Shoots field,
And achieve outstanding achievement in research.Up to the present, the method for multi-signal peptide analysis has not only been issued in field, and
There are many signal peptide analysis softwares to issue and be widely used.Signal peptide analysis method in field has stepped up signal peptide point
The precision of analysis, and attempt analysis understanding signal peptide from different angles.The conventional signal peptide analysis software of statistics, according to its progress
The forecast model applied during signal peptide analysis, they can be roughly divided into three classes by we：Signal peptide based on generation model is pre-
Survey method, the signal peptide prediction method based on discrimination model and the signal peptide prediction method based on sequence alignment.Based on differentiation mould
The signal peptide analysis method of type is typically that applied statistics analysis grader (as refreshing by network, SVMs etc.) carries out signal
Peptide analysis.The advantage of this method is that comparison is flexible, is adapted to the protein sequence of processing random length, has the disadvantage due to each ammonia
What base acid residue was coded separately between expression, each amino acid residue interrelated is just ignored.In addition, based on discrimination model
Signal peptide analysis method carry out cleavage site judgement when, easily true cleavage site the right and left be absorbed in local optimum fall into
Trap.Signal peptide analysis method based on generation model considers three functional areas (N areas, H areas, C areas) letter in signal peptide structure
Breath, such method passes through hidden Markov model (Hidden Markov by each interregional correlation of statistical signal peptide
Model, HMM) scheduling algorithm is modeled.Signal peptide analysis method advantage based on generation model is in the layer such as biology and chemistry
There is more intuitively interpretation on face, and explore correlation between signal peptide functional domain, this will be helpful to more deep understanding
The function and structure of these three functional areas；Have the disadvantage that being bad at process signal peptide has a case that longer amino acid sequence.Base
It is that the thoughts such as evolutionary conservatism and knowledge migration based on signal peptide are built in the signal peptide analysis method of sequence alignment, at this
In class signal peptide analysis method, it is necessary first to the data that signal peptide sequence composition includes annotation information are extracted from public database
Collection, then carries out sequence alignment by the signal peptide sequence in search sequence and data set.The advantage of this kind of signal peptide analysis method
It is in biology and physically to have stronger interpretation, the class model is more flexible compared to other two class models, Ke Yitong
Cross data set that dynamic more new model uses to update the signal peptide analysis model of issue, it is not necessary to the weight on new data set
New training pattern.
For signal peptide prediction, signal peptide and non-secreting egg can be distinguished well based on amino acid residue feature classifiers
White matter.But it is in hydrophobic amino acid fragment that signal peptide and N-terminal transbilayer helix fragment have one section in structure, so they
Amino acid residue feature has higher similitude.Therefore only with amino acid residue feature, grader is difficult to correctly recognize signal peptide
And transbilayer helix, so that the false positive of signal peptide prediction is too high.Signal peptide cutting site prediction is all a difficulty all the time
Point, reason is probably that signal peptide cutting site has complicated pattern, and signal peptide distribution of lengths is more disperseed.
The content of the invention
It is an object of the invention to for of the prior art not enough a kind of based on layering hybrid guided mode there is provided present invention proposition
The signal peptide and its cleavage site prediction algorithm of type, the algorithm introduce Functional domains letter for the first time in signal peptide prediction field
Cease to reduce signal peptide prediction after the false positive rate of N-terminal transbilayer helix, fusion amino acid residue and functional structure domain information
Performance is significantly improved.In terms of signal peptide cutting site prediction, the algorithm screens high-quality cleavage site by statistical rules and waited
Selected works, then calculate the sequence similarity fraction of candidate's cleavage site by sequence alignment, by merging statistical trust fraction
The final signal peptide cutting site with sequence similarity Score on Prediction.The present invention has relatively low vacation sun in terms of signal peptide prediction
Property, there is higher sensitivity in terms of signal peptide cutting site prediction.
Technical problem solved by the invention can be realized using following technical scheme：
Whether contain N using the SVM classifier identification of protein sequence based on amino acid residue feature in first layer first
Hold hydrophobic patches；Then applied in the second layer and be based on amino acid residue feature and the corresponding simple shellfish of functional structure characteristic of field
Ye Si and SVM classifier recognize that hydrophobic patches are signal peptide or N-terminal transbilayer helix；Finally in third layer, first according to statistics
Learning rules screening candidate's cleavage site and counting statistics trust fraction, then pass through Needleman-Wunsch sequences ratio again
The similarity scores of candidate signal peptide sequence are calculated algorithm, fraction is trusted statistics and sequence similarity fraction integration is determined
The signal peptide cutting site of prediction.
It is comprised the concrete steps that：
The first step：Using protein amino acid sequence extract PSSM matrix informations, secondary structure information, soluble information and
Amino acid physical chemistry information, feature extraction is carried out using mRMR, generates amino acid sequence feature.
Second step：The feature of protein sequence is extracted from CDD (Conserved Domain Database) database
Structure domain information, and functional-structural information is mapped to superfamily by the clustering information provided according to CDD, and occurred according to superfamily
Frequency generate frequent functional structure feature.
3rd step：SVM classifier is respectively trained according to two category features of protein sequence and Naive Bayes Classifier is used
In signal peptide prediction.
4th step：Protein sequence for being predicted as signal peptide, using asymmetric sliding window [- 13 ,+2] in albumen
Amino acid sequence segments are generated in matter sequence.
5th step：According to { -3, -1 ,+1 } of signal peptide rule, screen candidate's in the amino acid sequence segments of generation
Signal peptide cutting site, and calculate the corresponding statistical trust fraction of each candidate's cleavage site.
6th step：Corresponding signal peptide candidate sequence is generated by the signal peptide cutting site of candidate, and by the sequence and contained
The signal peptide sequence application Needleman-Wunsch algorithms for having annotation information do sequence alignment, and calculate per bars peptide candidate
The corresponding sequence similarity fraction of sequence and the signal peptide cutting site than being obtained by sequence pair.
7th step：Corresponding for each candidate's cleavage site statistical trust fraction and sequence similarity fraction are accumulated
Point, calculate the signal peptide cutting site finally predicted.
Compared with prior art, beneficial effects of the present invention are as follows：
(1) Model Fusion amino acid residue and functional nucleotide sequence structure domain information, significantly reduce the false positive of signal peptide prediction；
(2) the statistical fraction of Model Fusion and sequence similarity fraction, significantly improve signal peptide cutting site prediction
Sensitivity；
(3) from top and the model of lower leaf predicted signal peptide and its cleavage site, signal peptide and its cutting are significantly improved
The performance of site estimation.
Brief description of the drawings
Fig. 1 is the signal of the signal peptide of the present invention based on layering mixed model and its cleavage site Forecasting Methodology
Figure.
Fig. 2 is the signal peptide of the present invention based on layering mixed model and its output knot of cleavage site Forecasting Methodology
Fruit schematic diagram.
Embodiment
To be easy to understand technological means, creation characteristic, reached purpose and effect of the invention realized, with reference to
Embodiment, is expanded on further the present invention.
Referring to Fig. 1, signal peptide and its cleavage site Forecasting Methodology of the present invention based on layering mixed model, first
Whether contain N-terminal hydrophobic patches using the SVM classifier identification of protein sequence based on amino acid residue feature in first layer；
Then applied in the second layer and be based on amino acid residue feature and the corresponding naive Bayesian of functional structure characteristic of field and svm classifier
Device identification hydrophobic patches are signal peptide or N-terminal transbilayer helix；Finally in third layer, first waited according to statistical learning Rules Filtering
Select cleavage site and counting statistics trusts fraction, then calculate candidate by Needleman-Wunsch sequence alignment algorithms again
The similarity scores of signal peptide sequence, cut to the statistical signal peptide for trusting fraction and the integration determination prediction of sequence similarity fraction
Cut site.
Lower mask body is illustrated：
The first step：The corresponding evolution information (PSSM) of amino acid residue, structural information are extracted according to protein sequence information
(SS) and physical chemistry information (PI), each residue in such protein sequence can be represented with corresponding 30 dimensional vectors：
RF=[PSSM (20), SS (5), PI (5)] (1)
Second step：The corresponding feature of L amino acid of protein sequence N-terminal is taken, then application mRMR algorithms carry out feature choosing
Select the feature after generation optimization and be designated as sRF；
3rd step：Judge whether protein sequence includes hydrophobic patches (signal peptide using the SVM classifier based on sRF
With N-terminal transbilayer helix), terminate if not containing analysis；
4th step：The Functional domains for obtaining protein sequence using RPS-BLAST software retrieval CDD databases are believed
Breath, and generate energy property domains characteristic (FDF) according to the mapping relations of domain ID and domain superfamily；
FDFCDD=[δ1 δ2 … δi … δM] (2)
5th step：Judge that hydrophobic amino acid fragment is signal peptide or N using the Naive Bayes Classifier based on FDF
Transbilayer helix is held, the protein without FDF is predicted using the SVM classifier based on sRF features；
6th step：Cleavage site prediction is carried out to being predicted as the protein sequence comprising signal peptide, uses [- 13 ,+2] sliding
Dynamic window generation amino acid fragment, and root screens the cleavage point set of candidate in { -3, -1 ,+1 } rule and statistical learning algorithm；
Γ={ rk},ifθk＞ φT (5)
7th step：The sequence fragment that candidate's cleavage site generation according to filtering out is assumed, and by fragment and contain annotation
The protein sequence of information does sequence alignment, calculates similarity scores；
8th step：Corresponding to each candidate's cleavage site statistical trust fraction and sequence similarity fraction are accumulated
Point, it is the signal peptide cutting site finally predicted to choose the corresponding candidate's cleavage site of maximum.
Embodiment
An existing list entries, data are as follows：
>QuerySequence|SIGNAL 1 25
MIKSNRITACALAALFAGASFSASAWWGGPGYGNGLWDNMGDMFGDGYGDFNMSM
GGGGRGYGRGYGRGNGYGYGAPYGYGAPYGYGAPYGYGAPYGYGAPYGAMPYGA
MPPQMPAAPAQPQAAPSR
This is a sequence to be measured, as shown in Figure 2 using the software output result of the inventive method：
According to Signal-3L 2.0 engine,the predicted signal peptide is:1-
25
MIKSNRITACALAALFAGASFSASAWWGGPGYGNGLWDNMGDMFGDGYGDFNMSM
GGGGR GYGRGYGRGN
The protential cleavage sites and the credit scores
From the results, it was seen that this method is accurate and intuitively predicts signal peptide and its cleavage site.
The general principle and principal character and advantages of the present invention of the present invention has been shown and described above.The technology of the industry
Personnel are it should be appreciated that the present invention is not limited to the above embodiments, and the simply explanation described in above-described embodiment and specification is originally
The principle of invention, without departing from the spirit and scope of the present invention, various changes and modifications of the present invention are possible, these changes
Change and improvement all fall within the protetion scope of the claimed invention.The claimed scope of the invention by appending claims and its
Equivalent thereof.
Claims (4)
1. a kind of Forecasting Methodology of signal peptide and its cleavage site based on layering mixed model, it is characterised in that including as follows
Step：
1) whether N-terminal is contained using the SVM classifier identification of protein sequence based on amino acid residue feature in first layer first
Hydrophobic patches；
If 2) contain N-terminal hydrophobic patches, applied in the second layer and be based on amino acid residue feature and functional structure characteristic of field
Corresponding naive Bayesian and SVM classifier recognize that N-terminal hydrophobic patches are signal peptide or N-terminal transbilayer helix；
If 3) recognition result is signal peptide, first trusted according to statistical learning Rules Filtering candidate cleavage site and counting statistics
Fraction, then calculates the similarity scores of candidate signal peptide sequence by Needleman-Wunsch sequence alignment algorithms again, right
Statistics trusts fraction and sequence similarity fraction integration determines the signal peptide cutting site of prediction.
2. the Forecasting Methodology of signal peptide and its cleavage site according to claim 1 based on layering mixed model, it is special
Levy and be, the step 1) specific method be：
A. PSSM matrix informations, secondary structure information, soluble information and amino acid thing are extracted using protein amino acid sequence
Physicochemical information, feature extraction is carried out using mRMR, generates amino acid sequence feature.
B. the Functional domains letter of protein sequence is extracted from CDD (Conserved Domain Database) database
Breath, and functional-structural information is mapped to superfamily by the clustering information provided according to CDD, and the frequency occurred according to superfamily is given birth to
Into frequent functional structure feature.
3. the Forecasting Methodology of signal peptide and its cleavage site according to claim 2 based on layering mixed model, it is special
Levy and be, the step 2) specific method be：
C. SVM classifier and Naive Bayes Classifier is respectively trained according to two category features of protein sequence pre- for signal peptide
Survey.
4. the Forecasting Methodology of signal peptide and its cleavage site according to claim 3 based on layering mixed model, it is special
Levy and be, the step 3) specific method be：
D. for being predicted as the protein sequence of signal peptide, using asymmetric sliding window [- 13 ,+2] on protein sequence it is raw
Into amino acid sequence segments.
E. according to { -3, -1 ,+1 } of signal peptide rule, the signal peptide cutting of candidate is screened in the amino acid sequence segments of generation
Site, and calculate the corresponding statistical trust fraction of each candidate's cleavage site.
F. corresponding signal peptide candidate sequence is generated by the signal peptide cutting site of candidate, and by the sequence and contains annotation information
Signal peptide sequence application Needleman-Wunsch algorithms do sequence alignment, and calculate corresponding per bars peptide candidate sequence
Sequence similarity fraction and the signal peptide cutting site than being obtained by sequence pair.
G. it is integrated, is calculated most for each candidate's cleavage site corresponding statistical trust fraction and sequence similarity fraction
The signal peptide cutting site predicted eventually.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710140273.XA CN106951735B (en) | 2017-03-10 | 2017-03-10 | A kind of prediction technique of signal peptide and its cleavage site based on layering mixed model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710140273.XA CN106951735B (en) | 2017-03-10 | 2017-03-10 | A kind of prediction technique of signal peptide and its cleavage site based on layering mixed model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106951735A true CN106951735A (en) | 2017-07-14 |
| CN106951735B CN106951735B (en) | 2019-06-04 |
Family
ID=59467230
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710140273.XA Active CN106951735B (en) | 2017-03-10 | 2017-03-10 | A kind of prediction technique of signal peptide and its cleavage site based on layering mixed model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106951735B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110033822A (en) * | 2019-03-29 | 2019-07-19 | 华中科技大学 | Protein coding method and protein post-translational modification site estimation method and system |
| CN110970090A (en) * | 2019-11-18 | 2020-04-07 | 华中科技大学 | Method for judging similarity between polypeptide to be processed and positive data set peptide fragment |
| CN111091871A (en) * | 2019-12-19 | 2020-05-01 | 上海交通大学 | Fusion domain rule and deep learning based protein signal peptide and cleavage site prediction implementation method thereof |
| US10970604B2 (en) | 2018-09-27 | 2021-04-06 | Industrial Technology Research Institute | Fusion-based classifier, classification method, and classification system |
| CN115035947A (en) * | 2022-06-10 | 2022-09-09 | 水木未来(北京)科技有限公司 | Protein structure modeling method and device, electronic device and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2003014734A (en) * | 2001-06-15 | 2003-01-15 | Japan Science & Technology Corp | Method and computer program for discriminating signal peptide |
| CN1818916A (en) * | 2006-03-02 | 2006-08-16 | 上海交通大学 | Realization of kernel function predicted signal peptide and broken point based on sequence ratio |
| CN101175860A (en) * | 2005-03-15 | 2008-05-07 | 维莱尼姆公司 | Cellulases, nucleic acids encoding them and methods for making and using them |
| CN102012977A (en) * | 2010-12-21 | 2011-04-13 | 福建师范大学 | Signal peptide prediction method based on probabilistic neural network ensemble |
| CN102034029A (en) * | 2010-12-21 | 2011-04-27 | 福建师范大学 | Bayesian network-based signal peptide shearing site prediction method |
| CN106126972A (en) * | 2016-06-21 | 2016-11-16 | 哈尔滨工业大学 | A kind of level multi-tag sorting technique for protein function prediction |
-
2017
- 2017-03-10 CN CN201710140273.XA patent/CN106951735B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2003014734A (en) * | 2001-06-15 | 2003-01-15 | Japan Science & Technology Corp | Method and computer program for discriminating signal peptide |
| CN101175860A (en) * | 2005-03-15 | 2008-05-07 | 维莱尼姆公司 | Cellulases, nucleic acids encoding them and methods for making and using them |
| CN1818916A (en) * | 2006-03-02 | 2006-08-16 | 上海交通大学 | Realization of kernel function predicted signal peptide and broken point based on sequence ratio |
| CN102012977A (en) * | 2010-12-21 | 2011-04-13 | 福建师范大学 | Signal peptide prediction method based on probabilistic neural network ensemble |
| CN102034029A (en) * | 2010-12-21 | 2011-04-27 | 福建师范大学 | Bayesian network-based signal peptide shearing site prediction method |
| CN106126972A (en) * | 2016-06-21 | 2016-11-16 | 哈尔滨工业大学 | A kind of level multi-tag sorting technique for protein function prediction |
Non-Patent Citations (1)
| Title |
|---|
| YI-ZE ZHANG .ET AL: "Signal-3L 2.0: A Hierarchical Mixture Model for Enhancing Protein Signal Peptide Prediction by Incorporating Residue-Domain Cross-Level Features", 《JOURNAL OF CHEMICAL INFORMATION AND MODELING》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10970604B2 (en) | 2018-09-27 | 2021-04-06 | Industrial Technology Research Institute | Fusion-based classifier, classification method, and classification system |
| CN110033822A (en) * | 2019-03-29 | 2019-07-19 | 华中科技大学 | Protein coding method and protein post-translational modification site estimation method and system |
| CN110970090A (en) * | 2019-11-18 | 2020-04-07 | 华中科技大学 | Method for judging similarity between polypeptide to be processed and positive data set peptide fragment |
| CN111091871A (en) * | 2019-12-19 | 2020-05-01 | 上海交通大学 | Fusion domain rule and deep learning based protein signal peptide and cleavage site prediction implementation method thereof |
| CN111091871B (en) * | 2019-12-19 | 2022-02-18 | 上海交通大学 | Protein signal peptide and method for realizing prediction of cleavage site thereof |
| CN115035947A (en) * | 2022-06-10 | 2022-09-09 | 水木未来(北京)科技有限公司 | Protein structure modeling method and device, electronic device and storage medium |
| CN115035947B (en) * | 2022-06-10 | 2023-03-10 | 水木未来(北京)科技有限公司 | Protein structure modeling method and device, electronic device and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106951735B (en) | 2019-06-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106951735A (en) | A kind of Forecasting Methodology of signal peptide and its cleavage site based on layering mixed model | |
| Bansal et al. | A comparative analysis of K-nearest neighbor, genetic, support vector machine, decision tree, and long short term memory algorithms in machine learning | |
| Carmona et al. | NMEEF-SD: Non-dominated multiobjective evolutionary algorithm for extracting fuzzy rules in subgroup discovery | |
| CN109711464A (en) | Image Description Methods based on the building of stratification Attributed Relational Graps | |
| Ren et al. | A novel hybrid method of lithology identification based on k-means++ algorithm and fuzzy decision tree | |
| CN109241255A (en) | A kind of intension recognizing method based on deep learning | |
| CN106407333A (en) | Artificial intelligence-based spoken language query identification method and apparatus | |
| CN108804677A (en) | In conjunction with the deep learning question classification method and system of multi-layer attention mechanism | |
| CN110309723A (en) | A kind of driving behavior recognition methods based on characteristics of human body's disaggregated classification | |
| CN111611877B (en) | Anti-age-interference face recognition method based on multi-time-space information fusion | |
| CN109671102A (en) | A kind of composite type method for tracking target based on depth characteristic fusion convolutional neural networks | |
| CN106295245A (en) | The method of storehouse noise reduction own coding gene information feature extraction based on Caffe | |
| CN110135502A (en) | A kind of image fine granularity recognition methods based on intensified learning strategy | |
| CN105956529A (en) | Chinese sign language identification method based on LSTM type RNN | |
| CN105930663A (en) | Parkinson's disease early diagnosis method | |
| CN112215847B (en) | Method for automatically segmenting overlapped chromosomes based on counterstudy multi-scale features | |
| CN102779240B (en) | Inherent irregular protein structure forecasting method based on kernel canonical correlation analysis | |
| CN112163705A (en) | Epidemic situation prediction and early warning method based on CNN + SEIR and LSTM | |
| CN111123232B (en) | Radar individual identification system with task adaptability | |
| CN116432645A (en) | Traffic accident named entity recognition method based on pre-training model | |
| Ganapathiraju et al. | Transmembrane helix prediction using amino acid property features and latent semantic analysis | |
| CN116452688A (en) | Image description generation method based on common attention mechanism | |
| Zhang et al. | Improving protein secondary structure prediction by using the residue conformational classes | |
| CN114998731A (en) | Intelligent terminal navigation scene perception identification method | |
| Bagadi et al. | A robust feature selection method based on meta-heuristic optimization for speech emotion recognition |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |