Amino acid substitution matrices from protein blocks.

By Steven Henikoff, Jorja G. Henikoff. 1992

Subfields: Computational Biology & Bioinformatics

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computational Biology & Bioinformatics

Citations in the corpus

Improvement of molecular-replacement models with Sculptor (2011, positive): “Sculptor currently includes the identity, PAM250 (Dayhoff et al., 1978), BLOSUM50 and BLOSUM62 (Henikoff & Henikoff, 1992) matrices.”
BERTology Meets Biology: Interpreting Attention in Protein Language Models (2020, background): “One common substitution matrix is BLOSUM [29], which is derived from co-occurrence statistics of amino acids in aligned protein sequences.”
ProGen: Language Modeling for Protein Generation (2020, methodology): “BLOSUM62 is widely used across adopted alignment software (e.g., BLAST 2 ).”
Modeling aspects of the language of life through transfer-learning protein sequences (2019, background): “(vi) DeepBLOSUM65 (8 k free parameters): Each protein residue was encoded by its BLOSUM65 substitution matrix [98].”
DeepSF: deep convolutional neural network for mapping protein sequences to folds (2017, methodology): “The sequence alignment methods [11, 12] align the sequence of a target protein against the sequences of template proteins whose folds are known to generate alignment scores.”
ProDCoNN: Protein design using a convolutional neural network (2019, background): “percent of amino acid residues that are predicted the same as the native residue types or as a substitutable residue type corresponding to BLOSUM62 (scores from the matrix are positive).(63) The fuzzy”
ProtTrans: Towards Cracking the Language of Life’s Code Through Self-Supervised Deep Learning and High Performance Computing (2020, methodology): “Two approaches used amino acids instead of embeddings as input (marked as: no LMs : DeepOneHot [56] - one-hot encoding - and DeepBLOSUM62 [56] - input BLOSUM62 [69] substitution matrix), as well as, to the current state-of-the-art (SOA) method NetSurfP-2.”
A Survey on Metric Learning for Feature Vectors and Structured Data (2013, neutral): “For some applications, such as protein alignment or handwritten digit recognition, handtuned cost matrices may be available (Dayhoff et al., 1978; Henikoff and Henikoff, 1992; Micó and Oncina, 1998).”
Biological Sequence Modeling with Convolutional Kernel Networks (2017, positive): “We encode the sequences in Uniref50 using the BLOSUM62 position-independent probability matrix [Henikoﬀ and Henikoﬀ, 1992] by replacing each character with its corresponding substitution probability in BLOSUM62.”
BALBES: a molecular-replacement pipeline (2007, background): “4 ; ð1Þ where ‘score’ is based on the normalized BLOSUM62 substitution matrix (Henikoff & Henikoff, 1992), N1 and N2 are the number of residues in the first and the second sequence, Nalign is the number of aligned residues and Ngap is the number of gaps.”
A comprehensive review and performance evaluation of bioinformatics tools for HLA class I peptide-binding prediction (2020, background): “The sequence is first encoded into a matrix of the size 15 × 21 using BLOSUM62 as the encoding scheme.”
Tools for integrated sequence-structure analysis with UCSF Chimera (2006, background): “Alignment scores can include residue similarity, secondary structure information, and gap penalties: Residue similarity Any of several common substitution matrices (BLOSUM [11] or PAM [12]) can be used.”
Using deep learning to annotate the protein universe (2019, methodology): “Figure 5: (A) The amino acid embedding extracted from the trained ProtCNN model yields cosine similarities in embedding space that reflect the overall structure of the BLOSUM62 matrix [38].”
News from the protein mutability landscape. (2013, background): “(d) The predicted [15]); both known binding sites (positions 113 and 118) are score ranges from the most neutral (−100) to the strongest s “neutral” given the biophysical amino acid features as ne regions and in the BLOSUM62 matrix [17] for all other ge depicts intermediate scores (40 b score b 60),”