Terrier 0.4

.gitignore

@@ -186,4 +186,5 @@ drosophila-terrier.final.TEs.csv
 lightning_logs/
 comparison-test-data/get-rules.ipynb
 compare.sh
-drosophila.final.TEs.fa.gz
+drosophila.final.TEs.fa.gz
+terrier-0.4

README.rst

@@ -43,6 +43,17 @@ This model was trained using the Repbase library of repetitive DNA elements and
 
 An online version of Terrier (using CPUs only) is available at `https://portal.cpg.unimelb.edu.au/tools/terrier <https://portal.cpg.unimelb.edu.au/tools/terrier>`_.
 
+Latest Results
+==================================
+
+Terrier v0.4 was released and this was trained on RepBase31.04 (released 04-23-2026)
+which as substantially more sequences than the previous version (RepBase29.10)
+used to train the version of Terrier for the publication in Briefings in Bioinformatics (2025).
+
+This has altered the performance of the model. 
+
+See the `Latest Results <https://rbturnbull.github.io/terrier/latest/>`_ page for the performance of both versions of Terrier on the test data.
+
 Installation
 ==================================
 

comparison-test-data/build-results.py

@@ -1,3 +1,5 @@
+#!/usr/bin/env python3
+
 from terrier.evaluate import evaluate_metrics, confusion_matrix_fig, build_confusion_matrix, format_fig
 import plotly.graph_objects as go
 from plotly.subplots import make_subplots

comparison-test-data/comparison-results.csv

@@ -3,5 +3,5 @@
 1,TERL,0.9733333333333334,0.6301369863013698,0.9733333333333334,0.4657534246575342,0.7337367624810892,0.38762886597938145,0.659606656580938,0.16743119266055045,0.9187848729076256,0.694331983805668,0.78735275883447,0.6141732283464567,0.9055649241146712,0.7113594040968343,0.7835862844294548,0.6649928263988523
 2,TEclass2 (0.7),0.8133333333333334,0.8688524590163934,0.8133333333333334,0.819672131147541,0.6400602409638554,0.7858823529411765,0.6385542168674698,0.6839622641509434,0.815251084934904,0.8889733840304183,0.6776193428394296,0.9039341262580055,0.7886453063518831,0.8895224518888097,0.6942102304665543,0.9068825910931174
 3,TEclass2 (0.9),0.6666666666666666,0.86,0.6666666666666666,0.82,0.45331325301204817,0.8637873754152824,0.45180722891566266,0.7866666666666666,0.7383756974581525,0.9202350965575147,0.6236825790452573,0.9502982107355865,0.7048903878583473,0.9282296650717703,0.6290050590219224,0.9571045576407506
-4,Terrier (0.7),0.9466666666666667,0.9436619718309859,0.9066666666666666,0.9852941176470589,0.808095952023988,0.8794063079777366,0.6356821589205397,0.910377358490566,0.8527027027027027,0.8946117274167987,0.7587837837837837,0.9234194122885129,0.8462025316455696,0.9214659685863874,0.759493670886076,0.9583333333333334
-5,Terrier (0.9),0.9066666666666666,0.9705882352941176,0.8933333333333333,0.9850746268656716,0.6536731634182908,0.944954128440367,0.5367316341829086,0.9664804469273743,0.7797297297297298,0.9436741767764298,0.7128378378378378,0.9554502369668246,0.7854430379746835,0.9556809024979855,0.7221518987341772,0.9754601226993865
+4,Terrier (0.7),0.9733333333333334,0.9452054794520548,0.92,0.9855072463768116,0.815592203898051,0.8860294117647058,0.6596701649175413,0.9181818181818182,0.8418918918918918,0.8956661316211878,0.7304054054054054,0.9333950046253469,0.8424050632911393,0.9218632607062359,0.7481012658227848,0.955160744500846
+5,Terrier (0.9),0.9333333333333333,0.9857142857142858,0.9066666666666666,1.0,0.6731634182908546,0.9309576837416481,0.5367316341829086,0.9636871508379888,0.7655405405405405,0.9488084730803178,0.6817567567567567,0.9672943508424182,0.7664556962025316,0.9669694467382328,0.7018987341772152,0.9792605951307484

comparison-test-data/comparison-results.tex

@@ -6,8 +6,8 @@
 TERL & 97.3\% & 63.0\% & 97.3\% & 46.6\% & 73.4\% & 38.8\% & 66.0\% & 16.7\% \\
 TEclass2 (0.7) & 81.3\% & 86.9\% & 81.3\% & 82.0\% & 64.0\% & 78.6\% & 63.9\% & 68.4\% \\
 TEclass2 (0.9) & 66.7\% & 86.0\% & 66.7\% & 82.0\% & 45.3\% & 86.4\% & 45.2\% & 78.7\% \\
-Terrier (0.7) & 94.7\% & 94.4\% & 90.7\% & 98.5\% & 80.8\% & 87.9\% & 63.6\% & 91.0\% \\
-Terrier (0.9) & 90.7\% & 97.1\% & 89.3\% & 98.5\% & 65.4\% & 94.5\% & 53.7\% & 96.6\% \\
+Terrier (0.7) & 97.3\% & 94.5\% & 92.0\% & 98.6\% & 81.6\% & 88.6\% & 66.0\% & 91.8\% \\
+Terrier (0.9) & 93.3\% & 98.6\% & 90.7\% & 100.0\% & 67.3\% & 93.1\% & 53.7\% & 96.4\% \\
 \bottomrule
 \end{tabular}
 
@@ -20,7 +20,7 @@
 TERL & 91.9\% & 69.4\% & 78.7\% & 61.4\% & 90.6\% & 71.1\% & 78.4\% & 66.5\% \\
 TEclass2 (0.7) & 81.5\% & 88.9\% & 67.8\% & 90.4\% & 78.9\% & 89.0\% & 69.4\% & 90.7\% \\
 TEclass2 (0.9) & 73.8\% & 92.0\% & 62.4\% & 95.0\% & 70.5\% & 92.8\% & 62.9\% & 95.7\% \\
-Terrier (0.7) & 85.3\% & 89.5\% & 75.9\% & 92.3\% & 84.6\% & 92.1\% & 75.9\% & 95.8\% \\
-Terrier (0.9) & 78.0\% & 94.4\% & 71.3\% & 95.5\% & 78.5\% & 95.6\% & 72.2\% & 97.5\% \\
+Terrier (0.7) & 84.2\% & 89.6\% & 73.0\% & 93.3\% & 84.2\% & 92.2\% & 74.8\% & 95.5\% \\
+Terrier (0.9) & 76.6\% & 94.9\% & 68.2\% & 96.7\% & 76.6\% & 96.7\% & 70.2\% & 97.9\% \\
 \bottomrule
 \end{tabular}

docs/latest.rst

@@ -0,0 +1,72 @@
+
+Latest Results
+==================================
+
+Terrier v0.4 was released and this was trained on RepBase31.04 (released 04-23-2026)
+which as substantially more sequences than the previous version (RepBase29.10)
+used to train the version of Terrier for the publication in Briefings in Bioinformatics (2025).
+
+This has altered the performance of the model. See the tables below for the performance of both versions of Terrier on the test data.
+
+Rice
+-----
+
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+
+| Software            | Order Proportion Classified       | Order Accuracy       | Superfamily Proportion Classified        | Superfamily Accuracy         |
++=====================+===================================+======================+==========================================+==============================+
+| Terrier v0.3 (0.7) | 94.7%                             | 94.4%                | 90.7%                                    | 98.5%                        |
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+
+| Terrier v0.4 (0.7) | 97.3%                             | 94.5%                | 92.0%                                    | 98.6%                        |
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+
+| Terrier v0.3 (0.9) | 90.7%                             | 97.1%                | 89.3%                                    | 98.5%                        |
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+
+| Terrier v0.4 (0.9) | 93.3%                             | 98.6%                | 90.7%                                    | 100.0%                       |
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+
+
+
+Fruit Fly
+----------
+
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+
+| Software            | Order Proportion Classified       | Order Accuracy       | Superfamily Proportion Classified        | Superfamily Accuracy         |
++=====================+===================================+======================+==========================================+==============================+
+| Terrier v0.3 (0.7) | 80.8%                             | 87.9%                | 63.6%                                    | 91.0%                        |
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+
+| Terrier v0.4 (0.7) | 81.6%                             | 88.6%                | 66.0%                                    | 91.8%                        |
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+
+| Terrier v0.3 (0.9) | 65.4%                             | 94.5%                | 53.7%                                    | 96.6%                        |
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+
+| Terrier v0.4 (0.9) | 67.3%                             | 93.1%                | 53.7%                                    | 96.4%                        |
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+
+
+
+Human
+-----
+
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+
+| Software            | Order Proportion Classified       | Order Accuracy       | Superfamily Proportion Classified        | Superfamily Accuracy         |
++=====================+===================================+======================+==========================================+==============================+
+| Terrier v0.3 (0.7) | 85.3%                             | 89.5%                | 75.9%                                    | 92.3%                        |
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+
+| Terrier v0.4 (0.7) | 84.2%                             | 89.6%                | 73.0%                                    | 93.3%                        |
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+
+| Terrier v0.3 (0.9) | 78.0%                             | 94.4%                | 71.3%                                    | 95.5%                        |
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+
+| Terrier v0.4 (0.9) | 76.6%                             | 94.9%                | 68.2%                                    | 96.7%                        |
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+
+
+
+Mouse
+-----
+
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+
+| Software            | Order Proportion Classified       | Order Accuracy       | Superfamily Proportion Classified        | Superfamily Accuracy         |
++=====================+===================================+======================+==========================================+==============================+
+| Terrier v0.3 (0.7) | 84.6%                             | 92.1%                | 75.9%                                    | 95.8%                        |
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+
+| Terrier v0.4 (0.7) | 84.2%                             | 92.2%                | 74.8%                                    | 95.5%                        |
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+
+| Terrier v0.3 (0.9) | 78.5%                             | 95.6%                | 72.2%                                    | 97.5%                        |
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+
+| Terrier v0.4 (0.9) | 76.6%                             | 96.7%                | 70.2%                                    | 97.9%                        |
++---------------------+-----------------------------------+----------------------+------------------------------------------+------------------------------+