Web document 10.1. Syntax for sequences.

 

Source: from ftp://www.expasy.org/databases/prosite/release/profile.txt.

 

 

 

           A GENERALISED PROFILE SYNTAX FOR PROTEIN AND NUCLEIC ACID

                                SEQUENCE MOTIFS

 

                           Version 1.31, December 2001

 

 

 

   Philipp Bucher

   Swiss Institute of Bioinformatics (SIB)

   Institut Suisse de Recherches Experimentales sur le Cancer

   155 ch. des Boveresses, 1066 Epalinges s/Lausanne

   Switzerland

 

   Telephone: (+41 21) 692 58 92

   Electronic mail address: philipp.bucher@isb-sib.ch

 

 

   This document may be copied and redistributed freely, without advance per-

   mission, provided that this statement is reproduced with any copy.

 

 

 

                                  INTRODUCTION

 

 

   This document describes a general syntax to express a  quantitative,  pri-

   mary structure-based protein or nucleic acid sequence motif.  The designa-

   tion `quantitative' means that a motif description  assigns  a  degree  of

   similarity  to  a  potential  match rather than a binary status of true or

   false. The restriction `primary structure-based' implies that  the  proba-

   bility of finding a specific residue at one position is independent of any

   residue occurring at another position.

 

   The generalised profile syntax has been designed for and will be  used  in

   future releases of the PROSITE database.   In addition, it will be used in

   a similar database  of  nucleic  acid  sequence   motifs  currently  under

   development  by  the  author. Other researchers working on sequence motifs

   are encouraged to use the same format for their own motif collections, and

   may include this document in a public distribution release.

 

   The term `generalised profile syntax' is meant to indicate that  the  pro-

   posed  data  structure  represents  a  generalisation  of the profile type

   described by Gribskov et al. [1]. However, similar motif descriptors  have

   been  introduced by others under different names, e.g. weight matrices [2]

   or flexible patterns [3].

 

   The following terminology is adopted in this document:

 

   -  The term `profile' refers to a quantitative motif description based  on

      the generalised profile syntax.

   -  The term `pattern' refers to a qualitative motif description based on a

      regular  expression-like  syntax such as the one currently used in PRO-

      SITE entries marked as PATTERN.

   -  The term `motif' refers to the biological object one  attempts  to  ap-

      proximate by a pattern or a profile.

 

   Note that the  PROSITE  database  reserves  the  token  MATRIX to identify

   entries containing profiles.

 

                                     * * *

 

   The design of the new profile structure has been guided by various biolog-

   ical  and  technical  considerations.  High priority has been given to the

   following principles:

 

        A) Syntactic versatility

 

   The syntax should be versatile enough to cover a large variety of biologi-

   cally  relevant  motifs.  In particular, it should be be possible to accu-

   rately represent the following objects:

 

   -  Signatures for various types and levels of protein taxons.

   -  Highly degenerate protein structural and functional domains such as the

      immunoglobulin domains, or the SH2 and SH3 domains.

   -  Consensus sequences of interspersed repetitive DNA elements (SINEs  and

      LINEs).

   -  Basic gene expression signals, e.g. promoter elements,  RNA  processing

      signals, translational initiation sites.

   -  Recognition motifs of a large variety of sequence-specific  DNA-binding

      proteins.

   -  Protein and nucleic acid compositional domains, e.g. glutamine-rich ac-

      tivation domains, CpG islands.

 

 

        B) Determinative search instructions

 

   The profile syntax should have the capacity to encode precise and complete

   instructions  for  a  motif  search. Ideally, the result of a motif search

   should be determined by the profile and a sequence alone, i.e. not  depend

   on parameters of the search method. In practice, this goal may only be ap-

   proximately achieved due to ambiguities arising with multiple locally  op-

   timal profile-sequence alignments (see Section 2).

 

   Notes:

 

   -  In other implementations of similar methods, e.g. GCG profiles or HMMER

      Hidden  Markov  models software, different search methods are available

      as options and parameters of the search programs rather than as syntac-

      tic features of the motif description itself.  For the profiles in PRO-

      SITE, inclusion of determinative search instructions is a necessity be-

      cause  otherwise  the  information given on the NR lines (statistics of

      true and false positives/negatives) would have no meaning.

   -  The notion of determinative search instructions is not meant to imply a

      specific  search  algorithm.   There  is  space for different technical

      solutions to achieve the same result.

 

        C) Openness to different interpretations

 

   A profile syntax is situated at the interface between a  motif  definition

   and  a motif search method.  As such it can serve as a melting pot for in-

   tegrating complementary efforts.  While a rigid meaning vis-a-vis a search

   method  is  desirable, flexibility with regard to motif definition methods

   is equally important. In order to achieve such flexibility, it  is  essen-

   tial  that  the  profile  parameters remain open to a variety of different

   theoretical interpretations implicit in different methodologies.

 

   Relevant motif definition techniques may include both  comparative  struc-

   tural  and  wet  biochemical  approaches.  It is thus conceivable that the

   same type of numeric profile parameter may  reflect  log-probabilities  in

   one  case, and binding energies in another.  A profile syntax must be neu-

   tral in this respect in order to be generally acceptable to a  heterogene-

   ous  research  community relying on different rationales for motif defini-

   tion.

 

        D) Compatibility with existing search methods

 

   As a profile is required to encode determinative directives  for  a  motif

   search,  the underlying syntax should have the capacity to emulate most of

   the commonly used motif search techniques, such as:

 

   -  Search for PROSITE patterns.

   -  Search for fixed-length weight matrices without gaps [2].

   -  Search for complex motifs  defined  by  multiple  weight  matrices  and

      variable-length linkers [4].

   -  Gribskov's profile alignment algorithm [1].

   -  Barton's alignment algorithm for flexible patterns [3].

   -  Viterbi algorithm for the hidden Markov model architecture described in

      [5].

   -  The domain and fragment search algorithms implemented in the HMMER pro-

      grams hmmls and hmmfs, respectively [10].

 

   This requirement stems from two beliefs: (i) that the bewildering  variety

   of  motif search methods described in the literature can be understood and

   reformulated as special cases of a more general method; (ii) that such  an

   exercise  will  facilitate communication between different groups and will

   lead to new theoretical insights.

 

   Notes:

 

   -  With the exception of Barton's algorithm  for  flexible  patterns,  the

      capacity  of the generalised profile syntax to emulate the search tech-

      niques listed above has been verified by experiment.

 

                                     * * *

 

   The remaining part of this document is organised as follows:

 

   -  Section 1 explains the basic components of a profile which  are  likely

      to remain stable for several years.

   -  Section 2 presents accessory features necessary to encode complete  in-

      structions for a motif search. This part of the syntax may be gradually

      expanded in the future.

   -  Section 3 describes a specific machine-readable format  which  will  be

      used  in  PROSITE  and  a  similar  data  bank of nucleic acid sequence

      motifs.

   -  Section 4 shows two illustrative examples, one from  the  nucleic  acid

      and one from the protein world.

 

 

 

                    1) BASIC PROFILE STRUCTURE AND FUNCTION

 

   In abstract terms, a generalised profile can be described as an  alternat-

   ing  sequence  of  `match' and `insert' positions.  Match and insert posi-

   tions contain complementary sets  of  numeric  parameters  called  profile

   scores.  The  values assigned to these parameters may be different at each

   position.  In reality, a profile  resembles  a  two-dimensional  table  of

   numbers.

 

   From an other perspective, a profile may also be viewed  as  a  degenerate

   molecular sequence.  The match positions correspond to residues which typ-

   ically occur in such a sequence.  The insert  positions  represent  places

   where additional residues can optionally be inserted.

 

   The function of a profile is to align itself to a real sequence and to as-

   sign  a  number  to  such  an alignment.  This number is called similarity

   score or alignment score and serves to evaluate the significance of a  po-

   tential motif occurrence.

 

   The notion of a profile match cannot be separated from that of  an  align-

   ment.  The alignment is not only a prerequisite for computing a similarity

   score, it also expresses a specific interpration of  the  sequence  match.

   For  instance,  if  the  profile involved represent a protein domain where

   certain positions are associated  with  specific  functions,  e.g.   metal

   ion-binding  capacities  or  catalytic  roles, then the alignment will map

   these functions onto individual residues of the sequence.

 

   The basic components of a profile are those which are necessary  for  com-

   puting  a  similarity score. To prevent possible misunderstandings, it has

   to be stressed that a profile defines a score for any alignment, not  just

   for  an optimal alignment.  The concept of an optimal alignment relates to

   motif search strategies and is totally irrelevant in this Section.

 

 

   1.1) Definition of a profile-sequence alignment

 

   It is useful to introduce a profile-sequence alignment with the aid of the

   path  matrix  representation.   The following diagram defines an alignment

   between a sequence and a profile.

 

                            S   E   Q   U   E   N   C   E

                          .   .   .   .   .   .   .   .   .

                       p

                          .   .   .   .   .   .   .   .   .

                       r    \

                          .   . _ . _ .   .   .   .   .   .

                       o                \

                          .   .   .   .   .   .   .   .   .

                       f                    \

                          .   .   .   .   .   .   .   .   .

                       i                      |

                          .   .   .   .   .   .   .   .   .

                       l                        \

                          .   .   .   .   .   .   .   .   .

                       e

                          .   .   .   .   .   .   .   .   .

 

   The capital letters represent sequence residues,  the  lower-case  letters

   represent  profile  match  positions.   Profile  insert  positions are not

   marked by symbols.  They occur at the beginning, at the end,  and  between

   any pair of consecutive match positions of the profile.

 

   The path marked by horizontal, vertical, and  diagonal  bars  defines  the

   following alignment:

 

                                 S E Q U E - N

                                 r - - o f i l

 

   Such an alignment can also be defined by a sequence  of  path  matrix  co-

   ordinates.  By convention, the upper left corner of the matrix is assigned

   co-ordinates (0,0).  Note that path matrix co-ordinates correspond to pro-

   file  insert  positions  rather  than match positions. Likewise, they fall

   between consecutive residues of the sequence.  The above alignment is  de-

   fined by the following co-ordinate sequence.

 

         (1,0) , (2,1) , (2,2) , (2,3) , (3,4) , (4,5) , (5,5) , (6,6)

 

   In general, a co-ordinate sequence

 

   (i ,j ) , (i ,j ) , ... , (i ,j )

     0  0      1  1            L  L

 

   defines a valid sequence alignment between a profile of length N and a se-

   quence of length M if and only if:

 

       {      0 <= i  <= N  AND  0 <= j  <= M }  for  0 <= k <= L

                    k                  k

   AND

       {    ( i  + 1 = i    AND j  + 1 = j

               k        k+1      k        k+1

 

         OR ( i      = i    AND j  + 1 = j

               k        k+1      k        k+1

 

         OR ( i  + 1 = i    AND j      = j    }  for  0 <= k <= L-1 .

               k        k+1      k        k+1

   Notes:

 

   -  The above definition encompasses both global and local types of  align-

      ments.  In  the  following,  it is not necessary to distinguish between

      these two alternatives.  A global alignment may simply be viewed  as  a

      limit case of a local alignment.

   -  The alignment definition underlying generalised profiles is  equivalent

      to  the path definition of hidden Markon models in the following sense:

      (i) for a sequence and a profile of given lengths M,N,  the  number  of

      possible alignments is exactly identical to the number of paths through

      which an HMM of length N can generate sequences of length M, (ii) there

      is  an  obvious  one-to-one mapping between profile-sequence alignments

      and paths through HMMs; see also [9].

 

 

   1.2) Definition of the similarity score

 

   The similarity score of a profile sequence-alignment is  the  sum  of  the

   scores assigned to its scorable components.  The scorable components of an

   alignment are:

 

   -  the beginning

   -  each extension step

   -  each state transition

   -  the end

 

   Some of these terms need further explanation.

 

   An extension step occurs between any pair of consecutive path  matrix  co-

   ordinates.  There  are  three different types of extension steps: `match',

   `insert', and `deletion'  steps.  In  the  above  diagram,  diagonal  bars

   correspond  to match extension steps, horizontal bars correspond to insert

   extension steps, and vertical bars correspond to deletion extension steps.

   The number of extension steps defines the length of the alignment.

 

   The type of an extension step is also called a state. Each extension  step

   is thus associated with a match, insert, or deletion state.  At the begin-

   ning, an alignment is in `initiation' state.  At the end, it is in `termi-

   nation'  state.   Initiation,  match,  insert,  deletion,  and termination

   states will be symbolised by the letters B, M, I, D, and E, respectively.

 

   A state transition occurs between any two consecutive alignment components

   associated  with a state. Thus, there is one state transition for each co-

   ordinate pair of the alignment, including the first  and  the  last.  Note

   that  this  definition  implies  that state transitions also occur between

   identical states.

 

   In summary, an alignment of length L has:

 

      1  beginning

      L  extensions steps

    L+1  state transitions

      1  termination

   --------------------------------------------------------------------------

   2L+3  scorable components in total.

 

   All component scores are provided by the profile  in  a  position-specific

   manner.   Therefore, the similarity score does not depend on any parameter

   of an alignment method.  The types and functions of profile scores are now

   explained.

 

   The scores assigned to the beginning and end of the alignment  are  called

   `initiation'  and  `termination'  scores.  These  scores are distinct from

   those assigned  to  the  first  and  last  state  transition  though  they

   correspond  to  the same path matrix co-ordinates.  There are two types of

   scores for each class.  The `external' initiation  score  applies  to  co-

   ordinates  at  the  beginning  of the sequence.  The `internal' initiation

   score applies to all other co-ordinates.  External and  internal  termina-

   tion  scores  are defined analogously.  The function of these scores is to

   flexibly encode local or global alignment  scoring  modes.   In  addition,

   they  may  serve to anchor a motif at the beginning or at the end of a se-

   quence.

 

   The scores for extension steps comprise  three  classes:  match  extension

   scores,  insert extension scores, and deletion extension scores. Match and

   insert extension scores are  residue-specific  because  the  corresponding

   alignment steps span one sequence residue.  By contrast, there is only one

   deletion extension score per profile position because  deletion  steps  do

   not involve sequence residues.

 

   There are 16 different types of state transition scores for  all  possible

   transitions from an element of {B,M,I,D} to an element of {M,I,D,E}. State

   transition scores serve similar functions as gap opening  penalties  in  a

   sequence-sequence alignment.

 

 

   1.3) Basic profile structure

 

   The basic profile structure follows almost conclusively from the  forgoing

   definitions  of  a  profile-sequence  alignment  and its similarity score.

   What remains to be clarified are a few details.

 

   A profile is based on a particular alphabet.  The alphabet is considered a

   basic constituent of the profile because it determines the exact number of

   parameters per insert and match position.

 

   The two standard character sets for biomolecular patterns are:

 

   -  {A,C,G,T}                                      for nucleic acid motifs.

   -  {A,B,C,D,E,F,G,H,I,K,L,M,N,P,Q,R,S,T,V,W,Y,Z}  for protein  motifs.

 

   Other alphabets, e.g. alphabets including ambiguous codes for nucleotides,

   may be useful in particular circumstances.

 

   There is one insert and one match extension score for  each  character  of

   the  alphabet.   In practice, it is useful to define one additional insert

   and match extension score to deal with unexpected characters appearing  in

   real sequences.

 

   Some of the previously introduced profile scores are associated  with  in-

   sert  positions,  others  with match positions.  A look at the path matrix

   diagram makes clear which type of score is associated with which  type  of

   profile position.

 

   An insert position of a profile based on a K-letter alphabet contains  the

   following parameters:

 

      1  external initiation score

      1  internal initiation score

     16  state transition scores for all transitions between

         elements of {B,M,I,D} and {M,I,D,E}

      K  insert extension scores for each character of the alphabet

      1  insert extension score for an unexpected character

      1  internal termination score

      1  external termination score

   --------------------------------------------------------------------------

   K+21  insert position scores in total

 

   A match position of a profile based on a K-letter  alphabet  contains  the

   following parameters.

 

      K  match extension scores for each letter of the alphabet

      1  match extension score for an unexpected character

      1  deletion extension score

   --------------------------------------------------------------------------

   K+ 2  match position scores in total

 

   Admissible values for profile scores are any integer or real number plus a

   special value representing a forbidden alignment step.  This value will be

   called `low-value' and behaves like minus infinity in mathematical  opera-

   tions.

 

   A profile has also a defined topology, either linear  or  circular.   Most

   profiles  will  be  linear.   Circular profiles may represent motifs which

   consist of a variable number of tandemly  repeated  units.   Note  that  a

   linear profile begins and ends with an insert position.

 

   Notes:

 

   -  The above list of position-specific profile scores represents the  max-

      imum number of supported features. Real profiles derived with an exist-

      ing method will rarely use all of them.  Concise  representation  of  a

      profile  can be achieved through specification of appropriate defaults;

      see examples in Section 4.

   -  There is some redundancy in the implemented parameter set allowing  for

      alternative  representations of functionally equivalent profiles.  This

      freedom could be used for scaling profile scores in units related to  a

      particular mathematical or physical interpretation, e.g.  probabilities

      of a hidden Markov model or thermodynamic quantities.

   -  The above definition of a sequence alignment  assumes  linear  topology

      for  both  the profile and the sequence. Generalisation to circular to-

      pology is straightforward.  An alignment between a circular profile and

      a linear sequence, or between a linear profile and a circular sequence,

      corresponds to a path on a cylindrical surface.  An alignment between a

      circular  profile  and  a  circular sequence corresponds to a path on a

      torus.

 

 

                             2) PROFILE ACCESSORIES

 

   The primary purpose of a profile is to identify as  reliably  as  possible

   biologically  relevant  motif occurrences in new sequences. The basic pro-

   file structure described in the previous Section is not sufficient to  de-

   fine  a  rational  search  strategy  to  this  end.  The accessory profile

   features presented here fill this  gap.  Appropriately  interpreted,  they

   complement  the  position-specific profile scores to provide determinative

   instruction for a motif search.  In addition, they guide the  interpration

   of potential matches.

 

 

   2.1) Cut-off value

 

   For a profile and a sequence of typical lengths, there  is  a  very  large

   number of possible alignments.  At most a few of them will be biologically

   relevant. The function of a cut-off value is to a priori exclude  a  large

   number  of alignments from further consideration by a profile search algo-

   rithm. The fate of the remaining alignments with similarity scores greater

   than  or equal to the cut-off value depends on a specific disjointness de-

   finition applied; see below.

 

   An important aspect of a cut-off value is  that  it  gives  a  qualitative

   meaning  to a profile.  This is a prerequisite for statistics on false po-

   sitives and false negatives obtained in a database  search,  as  currently

   provided by PROSITE.

 

   In certain situations, it is useful to supply more than one cut-off value,

   partitioning the range of alignment scores into multiple areas.  The areas

   may correspond to different degrees of certainty, ranges  of  evolutionary

   distance, or levels of physiological activity.

 

 

   2.2) Score normalisation instructions

 

   The profile-alignment scores defined in the previous  Section  are  called

   raw  scores.  In  most  cases, they will not lend themselves to meaningful

   biological interpretations and will therefore not be very helpful  in  the

   interpretation  of  a  potential match.  In practice, one is interested in

   questions like: What is the probability of finding a match  of  a  certain

   score  in  a  random  sequence?  How does the similarity score relate to a

   measurable property of the biological object? The purpose of normalisation

   instructions  is  to convert the raw score into directly interpretable un-

   its.

 

   There may be multiple normalisation modes for the same profile,  each  one

   associated  with  a  different  mathematical,  physical, or biological in-

   terpretation; see examples in Section 4.

 

   Normalisation functions are required to preserve  the  ranking  of  scores

   pertaining to alternative alignments between the same profile and the same

   sequence.  However, since normalisation functions may depend  on  sequence

   parameters such as length and residue composition, they will generally not

   preserve the order of scores pertaining  to  matches  from  different  se-

   quences arising in a database search.

 

   Notes:

 

   -  Cut-off values may be defined in raw score units  or  normalised  score

      units.

   -  Programs may rely on normalised rather  than  raw  scores  for  various

      operation, e.g. sorting of accepted matches in a database search.

   -  An expanding list of normalisation functions is presented  in  Appendix

      B.

 

 

        2.3) Disjointness definitions

 

   There are situations where only a single best alignment and its similarity

   score  are  of  interest.  This arises for instance with a profile serving

   exclusively as a signature for a protein  family.   More  frequently,  the

   same  motif  may  occur  more  than once in a given sequence, and each oc-

   currence will be of interest.

 

   In the first case, the motif search problem is simple and can be solved by

   a  standard  optimal  alignment algorithm such as described in [1]. In the

   second case, the task is more difficult and needs to be explained in  more

   detail.

 

   At first glance, the problem seems to  be  to  find  all  profile-sequence

   alignments  with  similarity  scores  greater than or equal to the cut-off

   value.  However, such an approach would not yield useful results because a

   high  scoring alignment typically belongs to a large group of very similar

   alignments with comparable scores. Two members of such a group may  differ

   only  by  an  additional  extension  step at one end of one alignment.  In

   sequence-sequence  comparisons,  a  cluster  of  related   alignments   is

   represented  by  a single highest scoring member.  This seems a reasonable

   procedure for profiles too.

 

   As a second approximation, one could therefore define the task of  finding

   multiple profile matches in the same sequence as one of finding as many as

   possible, locally optimal, but mutually disjoint  alignments  with  scores

   greater than or equal to the cut-off value.  What is still missing in such

   a statement of the problem is a precise definition of disjointness  and  a

   tie-braking  rule  to  choose between equally high-scoring alignments. The

   former is of fundamental importance and needs to be addressed  here.   The

   latter  may  be  considered a property of a specific algorithm and thus is

   beyond the scope of this document.

 

   There are many possible ways to define  disjointness  of  two  alignments.

   The  algorithms  described for finding multiple locally optimal alignments

   between pairs of sequences consider two alignments disjoint if  they  have

   no  extension  step  in common [6,7].  The two alignments specified in the

   path matrix diagram below illustrate this notion of disjointness.

 

                            S   E   Q   U   E   N   C   E

                          .   .   .   .   .   .   .   .   .

                       p        \

                          .   .   . _ .   .   .   .   .   .

                       r    \           \

                          .   .   .   .   .   .   .   .   .

                       o      |             \

                          .   .   .   .   .   .   .   .   .

                       f        \               \

                          .   .   . _ .   .   .   . _ . _ .

                       i                \

                          .   .   .   .   .   .   .   .   .

                       l                    \

                          .   .   .   .   .   .   .   .   .

                       e                        \

                          .   .   .   .   .   .   .   .   .

 

 

   However, such a definition will not be adequate for many motif search  ap-

   plication  because  it allows the same sequence residue to be matched with

   different profile positions. Imagine the  case  of  a  protein  structural

   domain.   There,  it is inconceivable that the same residue simultaneously

   participates in the formation of two physically distinct domains,  occupy-

   ing different places within these domains.

 

   There may be no single disjointness definition adequate for all  kinds  of

   biological  sequence  motifs which can be characterised by a profile.  For

   this and other reasons, a specific notion of disjointness  is  viewed  and

   implemented  as  a profile-inherent property rather than a variable of the

   alignment method.  In some cases, a  particular  definition  may  even  be

   derived  from a measurable property of the biological object. The conclud-

   ing example illustrates this point.

 

   The DNA recognition site of mammalian transcription factor Sp1 is about 14

   bp  long and can be fairly accurately represented by a conventional weight

   matrix.  Experiments have shown that the minimal center-to-center distance

   for  two sites to be simultaneously occupied by two proteins is 10 bp. For

   a profile representing an Sp1 binding site, an appropriate  criterion  for

   disjointness  would  require  that  the sequence segments aligned with the

   central 10 bp region of the recognition motif do not overlap.

 

   Notes:

 

   -  The problems related to disjoint alignments are not  specific  to  pro-

      files.  They also occur with qualitative variable-length patterns based

      on a regular expression-like syntax.

   -  An expanding list of alternative disjointness definitions is  presented

      in Appendix A.

   -  The algorithms for multiple pairwise sequence alignments  described  in

      [6,7]  can  easily be adjusted to the disjointness definitions proposed

      in Appendix A.

   -  Another principle for parsing multiple matches between an HMM and a se-

      quence is implemented in the HMMER programs hmmls and hmmfs [10].

 

 

                     3) A MACHINE-READABLE TEXT FILE FORMAT

 

   This Section describes the format conventions used  in  the  PROSITE  data

   bank for representation of profiles

 

 

   3.1) General format of the MA line

 

   The current PROSITE database reserves the MA  line  code  for  information

   specific to matrix entries.

 

   A profile typically extends over many MA lines.  The general format  of  a

   block of consecutive MA lines is as follows:

 

   MA   /KEYWORD: parameter=value; parameter=value; ... ; /KEYWORD:

   MA   parameter=value; parameter=value; ... ; /KEYWORD: ...

 

   The text is substructured into so-called data blocks, each  one  beginning

   with  a  keyword followed by a list of parameter specifications.  Keywords

   identify different types of data blocks with characteristic parameter sub-

   sets.  The  keywords  at  the beginning of each data block are enclosed by

   slash on the left side and by colon on the right side.  Individual parame-

   ter  specifications are delimited by semicolon.  There is also a semicolon

   at the end of each data block containing at least one parameter specifica-

   tion.

 

   A single word, quoted string, or number must be contained within one line.

   Otherwise,  there  are  no  rules guiding the placement of text units onto

   physical lines. Within one block, the parameter specifications can  appear

   in any order.

 

   The following keywords define valid data block types:

 

   /GENERAL_SPEC:      General specifications.

   /DISJOINT:          Disjointness definition for multiple matches.

   /NORMALIZATION:     Score normalisation instructions.

   /CUT_OFF:           Recommended cut-off value(s).

   /DEFAULT:           Defaults for position specific profile parameters.

   /I:                 Profile insert position.

   /M:                 Profile match position.

 

 

   3.2) The formats of different data block types

 

        3.2.1) The GENERAL_SPEC data block

 

   The GENERAL_SPEC data block provides general information  about  the  pro-

   file.  It has the following format:

 

   /GENERAL_SPEC: ALPHABET=string;

                 [ LENGTH=length; TOPOLOGY=topology; BEGIN=begin; END=end ]

                 [ LOG_BASE=log_base; P0=p0; P=random_model ]

 

   where:

 

   -  `string' is a quoted character string defining the  character  set  for

      which,  and  the order in which, position-specific match and insert ex-

      tension scores are provided in subsequent M and I data blocks.

   -  `length' is the length of the profile defined as the  total  number  of

      match positions.

   -  `topology' is one of the alternative words LINEAR or CIRCULAR.

   -  `begin' is an integer indicating the match position withing the profile

      where  the  described biological object begins (implying that positions

      before `begin' characterise  contextual  constraints).   This  together

      with the END feature may be useful for profiles characterising biologi-

      cal objects such as transmembrane helices in proteins, or exons in gene

      sequences.   As  an  instruction to software, this parameter means that

      sequence residues matching profile positions before the  `begin'  posi-

      tion should not be reported as being part of the biological object.

   -  `end' is an integer indicating the match position  within  the  profile

      where  the biological object ends; see also remarks on previous parame-

      ter.

   -  `log_base' is the logarithmic base that should be used when translating

      the  generalised  profile  (back)  into an HMM, see [9].  Popular loga-

      rithmic bases for the representation of HMMs, null-models, substitution

      matrices, etc. are tabulated in APPENDIX C.

   -  `p0' is a real number between 0 and  1  defining  the  insert-to-insert

      state  transition probability of the null-model that should be used for

      translating the generalised profile (back) into an HMM; see  [9].  This

      parameter  defines  a  geometric  length distribution over the sequence

      space.

   -  `random-model' is a real  number,  or  comma  separated  list  of  real

      numbers,  defining the residue emission probabilities of the null-model

      that should be used for translating the generalised profile (back) into

      an  HMM;  see  [9]. These numbers are not required to sum to 1 and thus

      should be renormalised by programs on input.  In  PROSITE,  the  random

      model is usually given as percent amino acid frequencies.

 

   The GENERAL_SPEC data block is mandatory and precedes any DEFAULT, M or  I

   data block.

 

   Implicit defaults:

 

   -  TOPOLOGY=LINEAR;

 

   Example:

 

   MA   /GENERAL_SPEC: ALPHABET='ACGT';

 

   Notes:

 

   -  The optional LENGTH parameter is purely informative and redundant.  The

      actual  length  of the profile is given by the sequence of subsequent I

      and M data blocks.

 

 

        3.2.2) The DISJOINT data block

 

   The DISJOINT data block provides a definition of disjointness for multiple

   profile-sequence  alignments,  or indicates that only one globally optimal

   alignment is of interest.  It has the following format:

 

   /DISJOINT: DEFINITION=name; parameters;

 

   where:

 

   -  `name' is a word from a controlled vocabulary identifying  one  of  the

      supported disjoint definitions listed in Appendix A.

   -  `parameters' is a list of parameter specifications for the  correspond-

      ing  disjointness  definition. Note that different disjointness defini-

      tions depend on different parameter sets; see APPENDIX A.

 

   The DISJOINT data block is mandatory.

 

   Example:

 

   MA   /DISJOINT: DEFINITION=PROTECT; N1=12; N2=40;

 

   Notes:

 

   -  Some disjointness definitions are parameter-free.  In  this  case,  the

      list of parameter specifications is empty.

   -  The list  of  supported  disjoint  definitions  constitutes  a  dynamic

      feature  of  the  format.   New  functions  may be added in the future.

      Suggestions are welcome.

 

 

        3.2.3) The NORMALIZATION data block

 

   A NORMALIZATION data block describes a  specific  normalisation  mode  for

   alignment scores. It has the following format:

 

   /NORMALIZATION: FUNCTION=name; parameters;

                   [ MODE=mode-nr; PRIORITY=rank; TEXT=string; ]

 

   where:

 

   -  `name' is a word form a controlled vocabulary identifying  one  of  the

      supported normalisation functions listed in Appendix B.

   -  `parameters' is a list of parameter specifications providing values for

      all  parameters  of  the corresponding normalisation function listed in

      APPENDIX B.

   -  `mode-nr' is an integer by which the normalisation mode can be referred

      to in CUT_OFF data blocks.

   -  `rank' is an integer assigning a relative priority to the normalisation

      mode with regard to various pattern search operations.

   -  `string' is a quoted string describing a score normalised according  to

      this mode.

 

   NORMALIZATION data blocks are optional.  There may be  several  NORMALIZA-

   TION data blocks per profile.

 

   The optional MODE parameter either appears in all or in none of  the  NOR-

   MALIZATION  data blocks.  If specified, the mode numbers form a contiguous

   integer range starting with 1.  If not specified, mode numbers are implied

   by the order in which the NORMALIZATION data blocks appear in the profile.

 

   The optional PRIORITY parameter either appears in all or none of the  NOR-

   MALIZATION data blocks. If not specified, priorities are equal to the mode

   numbers. The lower the priority number, the higher  the  priority  of  the

   normalisation mode, and vice-versa.

 

   Example:

 

   MA   /NORMALIZATION: MODE=1; FUNCTION=LINEAR; TEXT='Homology Score';

   MA      R1=-90.558; R2=0.57225;

 

   Notes:

 

   -  Normalisation functions may, in addition to the  parameters  listed  in

      APPENDIX  B,  depend  on characteristics of the sequence such as length

      and residue composition.

 

 

        3.2.4) The CUT_OFF data block

 

   A CUT_OFF data block defines a cut-off level.  It has the  following  for-

   mat:

 

   /CUT_OFF: SCORE=rscore;

             [ LEVEL=level; TEXT=string; N_SCORE=nscore; MODE=mode-nr ]

 

   where:

 

   -  `rscore' is an integer defining the cut-off value in raw score units.

   -  `level' is an integer identifying a cut-off level.

   -  `string' is a quoted character string  characterising  profile  matches

      with  scores  greater  than or equal to the corresponding cut-off value

      (but lower than any higher cut-off value specified).

   -  `nscore' is a real number, or a comma separated list of  real  numbers,

      defining  the cut-off value(s) in normalised score units calculated ac-

      cording to the mode(s) identified by mode-nr.

   -  `mode-nr' is an integer, or a comma separated list of integers,  refer-

      ring  to  one  or  several normalisation modes defined in NORMALIZATION

      data blocks.

 

   The CUT_OFF data block for level 0 is mandatory.  There  may  be  multiple

   CUT_OFF data blocks, one for each level.

 

   The LEVEL parameter is optional for level 0.  All other levels are  speci-

   fied  explicitly.   The levels assigned to alternative cut-off values, in-

   cluding level 0, form a contiguous integer range.

 

   The N_SCORE and MODE parameters are either both present  or  both  absent.

   If present, they contain the same number of elements.

 

   Example:

 

   MA   /CUT_OFF: LEVEL=0; SCORE=237; N_SCORE=7.5; MODE=1;

 

   Notes:

 

   -  Cut-off values in raw score units may be used by programs which do  not

      support a given normalisation mode.

   -  The list of supported normalisation  functions  constitutes  a  dynamic

      feature  of  the  format.   New  functions  may be added in the future.

      Suggestions are welcome.

 

 

        3.2.5) The DEFAULT data block

 

   The DEFAULT data block redefines defaults  for  position-specific  profile

   parameters and has the following format:

 

   /DEFAULT: [ SY_I=char1; SY_M=char2; parameters; ]

 

   where:

 

   -  `char1' is a quoted character representing a profile insert position in

      a profile-sequence alignment.

   -  `char2' is a quoted character representing a profile match position  in

      a profile-sequence alignment.

   -  `parameters' is a list of  parameter  specifications  defining  default

      values for one or several of the profile scores listed in the tables at

      the end of this Section.

 

   DEFAULT data blocks are optional.  There  may  be  multiple  DEFAULT  data

   blocks per profile.

 

   Implicit defaults:

 

   -  SY_I='-'; SY_M='X';

 

   Example:

 

   MA   /DEFAULT: B0= *; B1= *; E0= *; E1= *;

 

   Notes:

 

   -  The first DEFAULT data block redefines the implicit defaults  given  in

      the  tables at the end of this Section.  Subsequent DEFAULT data blocks

      consecutively redefine each other.

   -  Asterisk represents low-value in the example; see next Subsection.

 

 

        3.2.6) The I and M data blocks

 

   The I and M data blocks contain position-specific profile scores  for  in-

   sert and match positions.  They have the following formats:

 

   /I: [ SY=char1; parameters; ]

   /M: [ SY=char2; parameters; ]

 

   where:

 

   -  `char1' is a quoted character representing  the  corresponding  profile

      insert position in a profile-sequence alignment.

   -  `char2' is a quoted character representing  the  corresponding  profile

      match position in a profile-sequence alignment.

   -  `parameters' is a list of parameter specifications assigning values  to

      one  or  several  of the position-specific profile scores listed in the

      tables at the end of this Section.

 

   The profile scores specified in I and M data blocks overwrite the  current

   default  values  set  by  a preceding DEFAULT data block or initialised as

   shown in the tables at the end of this Section.

 

   The values assigned to profile scores may be integers, reals, or low-value

   represented  by  an asterisk.  Most profile scores are assigned one value.

   The exceptions are the residue-specific insert and match extension scores.

   These scores can either be assigned one value or a comma separated list of

   values, one for each character of the alphabet. The correspondence between

   scores  and  characters  is  defined by the order in which the alphabet is

   presented in the GENERAL_SPEC data block. A single value is equivalent  to

   a list of identical values.

 

   Each I data block characterises one insert position of the profile.   Each

   M  data block characterises one match position of the profile.  The physi-

   cal order of the M and I data blocks defines  the  logical  order  of  the

   corresponding  profile  positions.  Default match and insert positions are

   not always  specified  explicitly.   This  requires  further  explanation.

   Remember  that a profile consists of an alternating sequence of insert and

   match positions, and that a linear profile starts and ends with an  insert

   position.   Default  I  or  M data block are implied wherever the physical

   order of I and M data blocks does not conform to these rules.

 

   Example:

 

   MA   /I: B0= 0; B1= 0; /M: M= 11, 3, 3, 4;

 

   In case the above line describes a complete linear profile, a default  in-

   sert position is implied at the end.  In case it describes a circular pro-

   file, no additional profile position is implied.

 

   Notes:

 

   -  There has been some debate (and no decision  so  far)  whether  profile

      scores  should  be  required  to  be integers.  In PROSITE, all profile

      scores are represented as integers  and  existing  software  supporting

      this format actually requires integer representation. Integer represen-

      tation has thus become a `de facto' standard.

   -  A linear normalisation function implicitly defines an integer  to  real

      conversion of profile scores; see protein example in Section 4.

 

 

   Profile scores of insert positions and implicit defaults:

 

   Name  Default  Parameter description

   --------------------------------------------------------------------------

   B0    B0= 0    External initiation score

   B1    B1= 0    Internal initiation score

   E0    E0= 0    External termination score

   E1    E1= 0    Internal termination score

 

   BM    BM= 0    State transition score from state B to M

   BI    BI= *    State transition score form state B to I

   BD    BD= *    State transition score from state B to D

   BE    BE= *    State transition score from state B to E

   MM    MM= 0    State transition score from state M to M

   MI    MI= *    State transition score from state M to I

   MD    MD= *    State transition score from state M to D

   ME    ME= 0    State transition score from state M to E

   IM    IM= *    State transition score from state I to M

   II    II= 0    State transition score from state I to I

   ID    ID= *    State transition score from state I to D

   IE    IE= *    State transition score from state I to E

   DM    DM= *    State transition score from state D to M

   DI    DI= *    State transition score from state D to I

   DD    DD= 0    State transition score from state D to D

   DE    DE= *    State transition score from state D to E

 

   I     I = 0    Insert extension score(s) for characters included in the

                     alphabet

   I0    I0= 0    Insert extension score for a character not included in the

                     alphabet

   --------------------------------------------------------------------------

 

 

   Profile parameters of match positions and implicit defaults:

 

   Name   Default  Parameter description

   --------------------------------------------------------------------------

   M      M = 0    Match extension score(s) for characters included in the

                      alphabet

   M0     M0= 0    Match extension score for a character not included in the

                      alphabet

   D      D = 0    Deletion extension score

   --------------------------------------------------------------------------

 

 

                                  4) EXAMPLES

 

 

   4.1) E. coli promoters

 

   The profile shown below describes the major class  of  E.  coli  promoters

   recognised  by  RNA  polymerase-sigma  factor 70. It is based on work pub-

   lished in [4] and emulates the functionality of the promoter  search  pro-

   gram TARGSEARCH.

 

 

MA   /GENERAL_SPEC: ALPHABET='ACGT';

MA   /DISJOINT: DEFINITION=PROTECT; N1=37; N2=42;

MA   /NORMALIZATION: MODE=1; FUNCTION=LINEAR; R1=-90.558; R2=0.57225;

MA      TEXT='Homology Score';

MA   /NORMALIZATION: MODE=2; FUNCTION=LINEAR; R1=-10.198; R2=0.06215;

MA      TEXT='Log KBk2';

MA   /CUT_OFF: LEVEL=0; SCORE=237; N_SCORE=45.0; MODE=1;

MA   /DEFAULT: B0=*; B1=*; E0=*; E1=*;

MA   /I: B0= 0; B1= 0;

MA   /M: M= 11, 3, 3, 4;

MA   /M: M=  8, 4, 2, 7;

MA   /M: M=  8, 2, 4, 7;

MA   /M: M=  7, 4, 2, 8;

MA   /M: M=  8, 4, 4, 5;

MA   /M: M=  7, 3, 5, 6;

MA   /M: M=  3, 5, 5, 8;

MA   /M: M=  5, 2, 5, 9;

MA   /M: M=  5, 8, 5, 3;

MA   /M: M=  0, 1, 2,17; SY='T';

MA   /M: M=  1, 1, 1,18; SY='T';

MA   /M: M=  0, 2,17, 2; SY='G';

MA   /M: M= 14, 3, 1, 4; SY='A';

MA   /M: M=  5,11, 2, 5; SY='C';

MA   /M: M=  9, 2, 3, 7; SY='A';

MA   /M: M=  5, 5, 3, 9;

MA      /M:/M:/M:/M:/M:/M:/M:/M:/M:

MA      /I: MD=0; MM=1;/I: DM=1;/I: DM=1;/I: DM=6;/I: DM=14;/I: DM=6;/I: DM=1;

MA   /M: M=  4, 5, 2,10;

MA   /M: M=  5, 4, 5, 6;

MA   /M: M=  3, 5, 5, 8;

MA   /M: M=  4, 4, 8, 5;

MA   /M: M=  4, 5, 7, 6;

MA   /M: M=  0, 2, 2,17; SY='T';

MA   /M: M= 20, 0, 0, 1; SY='A';

MA   /M: M=  5, 3, 3, 9; SY='T';

MA   /M: M= 12, 3, 3, 3; SY='A';

MA   /M: M= 11, 4, 3, 4; SY='A';

MA   /M: M=  0, 1, 0,20; SY='T';

MA   /M: M=  7, 2, 6, 6;

MA   /M: M=  4, 7, 5, 5;

MA   /M: M=  6, 6, 6, 4;

MA   /I: E0=0; E1= 0;

 

 

   The profile is substructured into four operationally distinct modules

   totaling 45 match positions.

 

   Position  Module

   --------------------------------------------------------------------------

     1-16    Weight matrix for the -35 region including the core hexamer  box

             TTGACA at pos. 10-15.

    17-25    Fixed length linker module encoded by 9 dummy  match  positions.

             This  module  is  defined  on  the  first of the two indented MA

             lines.

    26-31    Variable length linker scoring module encoded by  7  consecutive

             insert  positions.   This module is defined on the second of two

             indented MA lines.

   32 -45    Weight matrix for the -10 region including the core hexamer  box

             TATAAT at pos. 37-42.

   --------------------------------------------------------------------------

 

   The variable length linker scoring module defines  the  following  scoring

   scheme.

 

                        # of bp between core     score

                           hexamers boxes

 

                              15                   1

                              16                   6

                              17                  14

                              18                   6

                              19                   1

                              20                   1

                              21                   1

 

   These scores are achieved as follows.  Format-proprietary  default  values

   for state transition scores (not over-written by local defaults) make sure

   that  deletions  and  insertions  can  only  occur  at   positions   where

   corresponding  scores  are  explicitly  specified. Insertion gaps are thus

   generally forbidden.  A deletion gap can only be opened at  the  beginning

   of  the  linker  length scoring module and must be closed within or at the

   end of this module. A promoter with the maximal linker length of 21 can be

   aligned  without gap to the profile. In this case, the linker length score

   is provided by MM=1 at insert position 25. Promoters with  linker  lengths

   15  to  20  require  a deletion gap in their alignment to the profile. The

   corresponding scores are provided by DM=1, 1, 6, 14, 6, 1 at insert  posi-

   tions 26, 27, 28, 29, 30, 31, respectively.

 

   Notes:

 

   -  The default low-values assigned to parameters B0, B1, E0, E1,  together

      with  the  exceptions B0, B1 = 0 at the beginning and E0, E1 = 0 at the

      end of the profile, define a global  alignment  algorithm  with  endgap

      weighting in the profile but not in the sequence.

   -  Two normalisation modes are defined in this profile. The names  of  the

      corresponding  scores,  `Homology score' and `log KBk2', are taken from

      the original publication.  The parameters of the  second  normalisation

      function  were derived by a linear regression analysis between homology

      scores and enzyme selectivities (defined as log KBk2) of 31  transcrip-

      tionally assayed promoters.

   -  The cut-off homology score of 45 has been proposed by  the  authors  as

      lower limit for effective promoters.

   -  The disjointness definition protecting only the TATAAT box region  from

      sequence  overlap,  is  motivated by a known promoter example where two

      adjacent TATAAT boxes direct transcription from two distinct initiation

      sites six bp apart from each other.

   -  This profile is not supposed to represent the  most  accurate  E.  coli

      promoter prediction method available today.  It primarily serves to il-

      lustrate that the proposed syntax is flexible  enough  to  express  the

      functionality of a specialised search algorithm developed for a partic-

      ular object.

 

 

   4.2) Src homology domain SH3

 

   The profile shown on the next page describes the Src homology  domain  SH3

   as  defined  by sequence similarity. It has been constructed by a recently

   described extension of Gribskov's method  incorporating  several  improve-

   ments [8].

 

   The SH3 profile consists of three homology blocks separated by two gap re-

   gions.  Within the homology blocks, small insertions and deletions are not

   totally forbidden but strongly impeded by high gap costs  defined  in  the

   DEFAULT  data  block: MI=-26, I=-3, MD=-26, D=-3.  These numbers are over-

   written by more permissive values in the two gap regions.

 

   Notes:

 

   -  The SH3 profile uses only features which are compatible with Gribskov's

      methodology.  As a consequence, it can be automatically reformatted for

      use with the existing profile alignment programs implemented in the GCG

      package.

   -  The second normalisation mode defines a real number conversion  of  the

      integer profile scores.

 

 

MA   /GENERAL_SPEC: ALPHABET='ACDEFGHIKLMNPQRSTVWY';

MA   /DISJOINT: DEFINITION=PROTECT; N1=1; N2=53;

MA   /NORMALIZATION: MODE=1; FUNCTION=GLE_ZSCORE; R1=44.55; R2=-0.0035;

MA      R3=0.7386; R4=1.001; R5=0.208; TEXT='ZScore';

MA   /NORMALIZATION: MODE=2; FUNCTION=LINEAR; R1=0.0; R2=0.1;

MA      TEXT='OrigScore';

MA   /CUT_OFF: LEVEL=0; SCORE=90; N_SCORE=7.0; MODE=1;

MA   /DEFAULT: MI=-26; I=-3; IM=0; MD=-26; D=-3; DM=0;

MA   /M: SY='F';M=-2,-3,-3,-4,2,-3,-2,1,-2,0,-1,-2,-3,-3,-4,-2,-1,0,-5,2;

MA   /M: SY='I';M=-1,-5,-2,-3,-2,-3,0,1,1,-1,1,-1,-2,-1,1,-1,0,1,-4,-4;

MA   /M: SY='A';M=2,-3,1,0,-5,2,-2,-1,-1,-3,-2,1,1,0,-2,2,2,0,-8,-5;

MA   /M: SY='L';M=-3,-8,-5,-4,2,-6,-2,2,-4,6,4,-3,-3,-2,-3,-3,-2,1,-3,0;

MA   /M: SY='Y';M=-4,-2,-6,-6,9,-7,0,-1,-5,-1,-3,-3,-6,-5,-6,-4,-4,-4,-1,11;

MA   /M: SY='D';M=1,-6,3,3,-7,0,0,-2,-1,-4,-3,2,0,1,-2,0,0,-2,-9,-6;

MA   /M: SY='Y';M=-5,-3,-6,-6,10,-7,-1,-1,-2,-1,-2,-3,-6,-5,-5,-4,-4,-4,-1,11;

MA   /M: SY='K';M=-1,-6,1,1,-4,-2,0,-2,2,-3,-1,1,-1,1,1,0,0,-3,-7,-6;

MA   /M: SY='A';M=1,-4,1,0,-5,1,-1,-1,0,-3,-1,1,0,0,0,1,1,-1,-7,-6;

MA   /M: SY='R';M=0,-5,0,0,-5,-1,0,-1,1,-3,-1,1,0,1,1,0,0,-2,-5,-5;

MA   /M: SY='R';M=0,-5,1,1,-6,0,1,-2,1,-4,-2,1,0,1,2,1,0,-2,-5,-5;

MA   /M: SY='E';M=1,-6,2,2,-6,0,0,-2,-1,-4,-2,1,1,1,-1,0,0,-3,-8,-6;

MA   /M: SY='D';M=0,-6,2,2,-6,0,1,-3,0,-5,-3,2,-1,2,-1,0,0,-4,-7,-4;

MA   /M: SY='D';M=0,-8,4,3,-6,0,0,-2,-1,-3,-2,2,-2,2,-2,0,-1,-3,-9,-6;

MA   /M: SY='L';M=-2,-8,-5,-5,2,-5,-3,3,-4,7,5,-4,-3,-3,-4,-3,-2,3,-4,-2;

MA   /M: SY='S';M=1,-4,1,1,-5,1,0,-2,1,-4,-2,1,0,0,0,1,1,-2,-6,-5;

MA   /M: SY='F';M=-3,-7,-6,-6,6,-5,-3,3,-2,5,3,-4,-5,-4,-5,-4,-3,1,-3,3;

MA   /M: SY='Q';M=-1,-6,0,0,-3,-2,1,-1,1,-2,0,0,-1,1,1,-1,0,-1,-6,-4;

MA   /M: SY='K';M=-1,-8,0,1,-3,-2,0,-2,3,-3,0,1,0,2,2,0,0,-3,-6,-6;

MA   /M: SY='G';M=2,-5,1,0,-7,7,-3,-4,-2,-6,-4,1,-1,-2,-4,2,0,-2,-10,-8;

MA   /M: SY='D';M=1,-7,5,4,-8,1,1,-3,0,-5,-3,2,-1,2,-2,0,0,-4,-10,-6;

MA   /M: SY='I';M=0,-5,-1,-2,-2,-2,-1,2,0,0,1,-1,-2,0,0,-1,0,1,-6,-5;

MA   /M: SY='L';M=-2,-6,-5,-5,3,-5,-3,4,-3,6,4,-4,-4,-3,-4,-3,-2,3,-5,0;

MA   /M: SY='Q';M=-1,-5,-1,-1,-3,-2,0,0,0,-2,-1,0,-1,0,0,-1,0,-1,-6,-3;

MA   /M: SY='V';M=0,-4,-3,-4,-1,-3,-3,5,-3,3,3,-2,-2,-2,-3,-2,0,5,-8,-4;

MA   /M: SY='L';M=-1,-6,-3,-3,-1,-3,-2,2,-3,3,2,-2,-2,-2,-3,-2,-1,2,-5,-3;

MA   /M: SY='D';M=0,-6,3,3,-6,0,1,-3,2,-5,-2,2,-1,2,1,0,0,-4,-7,-5;

MA   /M: SY='K';M=-1,-6,0,0,-2,-1,0,-3,3,-4,-1,1,-1,0,1,0,0,-3,-6,-4;

MA   /M: SY='N';M=1,-4,1,1,-5,0,0,-2,0,-3,-2,1,1,0,-1,1,1,-1,-7,-5;

MA      /I: MI=0; I=-1; MD=0; /M: SY='X'; M=0; D=-1;

MA   /M: SY='G';M=1,-5,0,0,-5,1,-2,-1,-2,-3,-2,0,0,-1,-2,0,0,-1,-8,-6;

MA   /M: SY='G';M=1,-6,3,3,-7,3,0,-4,-1,-5,-4,2,-1,1,-2,1,0,-3,-10,-6;

MA   /M: SY='W';M=-9,-12,-9,-11,1,-11,-4,-8,-5,-3,-6,-6,-8,-7,3,-4,-8,-9,26,0;

MA   /M: SY='W';M=-7,-9,-9,-9,0,-9,-4,-5,-5,-1,-4,-6,-7,-6,2,-3,-6,-6,18,-1;

MA   /M: SY='K';M=-1,-7,0,0,-3,-2,0,-2,2,-3,-1,1,-1,1,2,0,-1,-3,-5,-5;

MA   /M: SY='G';M=2,-3,0,-1,-6,3,-3,-2,-3,-4,-3,0,0,-2,-3,1,0,0,-10,-6;

MA   /M: SY='Q';M=-2,-6,0,0,-3,-3,1,-2,0,-2,-1,0,-2,1,1,-1,-1,-3,-5,-3;

MA      /I: MI=0; I=-2; MD=0; /M: SY='X'; M=0; D=-2;

MA   /M: SY='T';M=0,-4,-1,-1,-4,0,-2,0,-1,-2,0,0,-1,-1,-1,0,1,0,-7,-5;

MA   /M: SY='T';M=0,-5,0,0,-3,-1,-1,-1,1,-3,-1,1,-1,0,0,1,1,-1,-6,-4;

MA   /M: SY='G';M=0,-5,0,-1,-5,3,-2,-3,-1,-5,-3,0,-1,-1,-1,1,0,-2,-7,-6;

MA   /M: SY='K';M=0,-6,1,1,-5,-1,1,-2,2,-4,-1,1,-1,2,2,0,0,-3,-6,-6;

MA   /M: SY='R';M=-1,-6,-1,-1,-5,-3,1,-1,1,-3,-1,0,-1,1,3,-1,-1,-2,-2,-6;

MA   /M: SY='G';M=1,-5,0,0,-6,6,-3,-3,-3,-5,-4,0,-1,-2,-4,1,0,-2,-10,-6;

MA   /M: SY='W';M=-5,-5,-5,-5,2,-6,-2,-2,-4,-1,-3,-3,-6,-5,-3,-3,-4,-4,4,3;

MA   /M: SY='F';M=-3,-5,-6,-6,6,-5,-3,4,-1,3,2,-4,-4,-5,-4,-3,-2,2,-4,3;

MA   /M: SY='P';M=2,-4,-1,-1,-7,-1,0,-3,-2,-4,-3,-1,8,0,0,1,0,-2,-8,-7;

MA   /M: SY='G';M=1,-3,0,0,-4,2,-1,-2,0,-3,-2,0,0,-1,-1,1,1,-1,-6,-5;

MA   /M: SY='N';M=1,-5,2,1,-5,0,1,-2,1,-4,-2,2,0,0,0,1,1,-2,-7,-4;

MA   /M: SY='Y';M=-5,-1,-7,-7,10,-8,-1,-1,-5,-1,-3,-3,-7,-6,-6,-4,-4,-5,0,13;

MA   /M: SY='V';M=0,-3,-3,-5,-2,-2,-3,5,-3,2,2,-2,-2,-3,-4,-1,0,5,-8,-5;

MA   /M: SY='E';M=1,-6,2,3,-6,0,0,-2,1,-4,-2,1,0,2,0,0,0,-3,-8,-6;

MA   /M: SY='P';M=0,-5,-1,-1,-2,-2,-1,-2,-1,-3,-2,0,1,-1,-2,0,-1,-2,-6,-3;

 

 

   Acknowledgements:

 

   I thank Roland Luethy, Michael Gribskov, Stephen Altschul, David Haussler,

   Sean  Eddy,  Kevin Karplus, and Ewan Birney for valuable comments and dis-

   cussions.  The text file format described in Section 3 has  been  designed

   in  collaboration with Amos Bairoch.  Ioannis Xenarios has contributed the

   SH3 profile shown with minor modifications in Section 4.

 

 

                                   REFERENCES

 

 

   1  Gribskov M., Luethy R., Eisenberg D.

      Profile analysis.

      Meth. Enzymol. 183:146-159(1990).

 

   2  Staden R.

      Searching for patterns in protein and nucleic acid sequences.

      Meth. Enzymol. 183:193-211(1990).

 

   3  Barton G.J., Sternberg M.J.E.

      Flexible protein sequence patterns: a sensitive method to detect weak

      structural similarities.

      J. Mol. Biol. 212:389-402(1990).

 

   4  Mulligan M.E., Hawley D.K., Entriken R., McClure W.R.

      Escherichia coli promoter sequences predict in vitro RNA polymerase

      selectivity.

      Nucleic Acids Res. 12:789-800(1984).

 

   5  Krogh A., Brown M., Mian I.S., Sjoelander K., Haussler D.

      Hidden Markov models in computational biology.

      J. Mol. Biol. 235:1501-1531(1994).

 

   6  Waterman M.S., Eggert M.

      A new algorithm for best subsequence alignments with application to

      tRNA-rRNA comparisons.

      J. Mol. Biol. 197:723-728(1987).

 

   7  Huang X., Miller W.

      A time-efficient, linear-space local similarity algorithm.

      Adv. Appl. Math. 12:337-357(1991).

 

   8  Luethy R., Xenarios I., Bucher P.

      Improving the sensitivity of the sequence profile method.

      Protein Sci. 3:139-146(1994).

 

   9  Bucher P., Karplus K., Moeri N., Hofmann K.

      A flexible motif search technique based on generalized profiles.

      Comput. Chem. 20:3-24(1996).

 

  10  Eddy S.R.

      Hidden Markov models.

      Curr. Opin. in Struct. Biol. 6:361-365(1996).

 

 

 

                     APPENDIX A): DISJOINTNESS DEFINITIONS

 

   The following notions of disjointness of two alignments between  the  same

   profile and the same sequence are currently defined:

 

   Name    Parameters  Description

   --------------------------------------------------------------------------

   UNIQUE              Multiple profile-sequence alignment between  the  same

                       profile  and the same sequence are not permitted.  The

                       result of a profile search consists of a  single  best

                       alignment.

 

   PROTECT N1 (int)    Two profile-sequence  alignments  are disjoint  if the

           N2 (int)    two sequence segments  associated with the `protected'

                       profile  area  do  not  overlap. The protected profile

                       area extends from match position N1 to match  position

                       N2 inclusive.

   --------------------------------------------------------------------------

 

 

                     APPENDIX B): NORMALIZATION FUNCTIONS

 

   The following score normalisation functions are currently defined.

 

   Name        Parameters  Formula (X = raw score, Y= normalised score)

   --------------------------------------------------------------------------

   LINEAR      R1 (real)   Y = R1 + R2 * X

               R2 (real)

 

   GLE_ZSCORE  R1 (real)       X/[R1*(1.0-exp(R2*SeqLen-R3))]-R4

               R2 (real)   Y = ---------------------------------

               R3 (real)                       R5

               R4 (real)

               R5 (real)

   --------------------------------------------------------------------------

 

   Sequence-dependent variables:

 

   -  SeqLen (integer) is the length of the sequence.

 

   Notes:

 

   -  The score normalisation function named GLE_ZSCORE has been described by

      Gribskov, Luethy, and Eisenberg in [1].

 

 

                 APPENDIX C): FREQUENTLY USED LOGARITHMIC BASES

 

   Logarithmic base    Name of units        Used e.g. in:

   --------------------------------------------------------------------------

   10.000000000        Log10 units          PROSITE profile -log P-values

    2.718281828        nats                 SAM HMM log-odds scores

    2.000000000        bits                 BLAST, HMMER HMM log-odds scores

    1.414213562        1/2 bit units        BLAST blosum62 matrix

    1.259921050        1/3 bit units        BLAST blosum45, pam250 matrices

    1.258925412        1/10 Log10 units     Dayhoff MDM78 matrix

    1.189207115        1/4 bit units        BLAST blosum35, blosum40 matrices

    1.148698355        1/5 bit units        BLAST blosum30 matrix

    1.000693387        1/1000 bit units     HMMER integer arithmetics

   --------------------------------------------------------------------------